Comaty MA 2013

eeh power systems
laboratory
Farid Comaty
Modeling and Simulation of the European

Power System using Power Nodes -
Assessing the Value of Flexibility for High-Share
Integration of Renewable Energies in Europe
Master Thesis
PSL1216
EEH – Power Systems Laboratory

Swiss Federal Institute of Technology (ETH) Zürich
Expert: Prof. Dr. Göran Andersson

Supervisor: Dipl.-Ing. Andreas Ulbig
Zürich, August 31, 2013

ii
"Our dependence on fossil fuels amounts to global pyromania, and the only fire
extinguisher we have at our disposal is renewable energy."
Hermann Scheer
April 29, 1944 - October 14, 2010
Social Democrat Member of the German Parliament
Abstract
This master thesis assesses the value of flexibility of the European power sys-
tem (EPS) in order to derive an optimal operation strategy for accommodating high
shares of solar and wind energy. It also investigates the related policy, economic
and environmental implications of renewable energy sources (RES) and discusses
country specific flexibility value with a focus on storage units.
The EPS is modeled as a system of interconnected power nodes, each one
representing a country’s generation and consumption portfolio. Data basis are IEA
and ENTSO-E projections for 2020 and EurElectric hydro data. The wind and
solar hourly time series were obtained from the IRENE-40 EU research project.
The flexibility sources analyzed are pump storage plants (power and energy rating),
power lines net transmission capacity (NTC) and curtailment of renewable energy
sources. The simulation length is one full year and the system is dispatched hourly
based on the merit order curve. A model predictive optimization algorithm looks
24 hour forward and updates the dispatch strategy every four hours to optimally
cycle storage reservoirs.
Moving from a low-RES (20%) to a high-RES scenario (70%) where wind and
solar energy contributes up to 53%, the EPS saves on total operation cost in gener-
ating electricity between €∼58 to 95 bn, depending weather the O&M cost of vari-
able renewable plants is considered. In such scenario, the priority in-feed policy
for RES units is no longer economically optimal; spilling up to 16% of available
free energy saves €bn 8bn on operation cost, thereby valuating the curtailment
of RES. To minimize curtailments, the NTC capacities and pump storage ratings
(power and energy) should best be tripled. Above such increase, curtailments will
still be necessary in a high-share RES scenario and the flexibility value starts to
saturate, meaning that an increase in the flexibility of the system is no longer help-
ful for integrating more RES. On a European system-level, RES integration was
most sensitive to increasing the net transmission capacity between the countries
with an almost linear slope of two percentage points, while for storage power rat-
ing the slope of integration was one percentage point and for storage energy rating
the slope of integration was half a percentage point. Tripling each flexibility source
saves on operation cost €bn 15.5, 2.7 and 0.66, respectively, and €bn 18.42 if all
were tripled simultaneously. The yearly savings of operation cost represents the
flexibility value. CO2 emissions from the EU power sector are cut by 91% only
iv
v
if coal and oil power plants are replaced by natural gas or equipped with carbon
capture and storage technology, otherwise by a maximum of 44%.
The thesis discusses also country case studies which did not follow the general
EU trend, examples are Spain and Germany, and suggests specific power system
upgrades accordingly. As well, we illustrate a forecast of the power exchange pat-
terns between EU countries under a high renewable energy scenario and predict
the future cost of electricity generation and market price. All different power sys-
tem setups were adequate as no load shedding occurred which was above 0.5 % of
national load demand.
Acknowledgments
This work was conducted at the Power Systems Laboratory (PSL) of ETH
Zurich and in collaboration with the department of "Future Technology Execution"
of Alstom (Schweiz) AG located in Birr.
Analyzing the numerous aspects of a high-renewable electricity supply in 29
countries was a very challenging task, which I could have not undertake without
the commitment of Andreas Ulbig to supervise my thesis. I pose in deep gratitude
towards him for allowing me to work on such an interesting topic, helping me
figure out the bottlenecks in the thesis and deepen my knowledge on power system
operation. Also, I sincerely thank Fabio Feretti and Roberto Bove from Alstom
for their continuous support and the constructive brainstorming sessions we had to
structure the problem and arrive to a final message. I am as well grateful to Prof.
Dr. Andersson for being my tutor in my Master studies at ETH and allowing me to
write my thesis at PSL.
My post-graduate studies would have not been that enjoyable without my dear
friends Abhishek, Cooper, Erik, June, Martin, Moonjo and Ratri, whom I have met
here. Special thanks to Erik for his helpful advices on structuring this report.
Finally, I wish to sincerely thank the two pillars of my education in life, my
supportive parents Nicolas and Joumana Comaty, whom have always encouraged
me to build my own path and work hard for it. I am grateful also to the Govern-
ment of Quebec in Canada for supporting part of my undergraduate and graduate
studies.
vi
Contents
List of Abbreviations xi
List of Figures xi
List of Tables xiii
1 Introduction 2
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Sources of operational flexibility in the grid . . . . . . . . . . . . 6
1.2.1 Ramping capability of generators . . . . . . . . . . . . . 6
1.2.2 Energy storage . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.3 Demand side participation . . . . . . . . . . . . . . . . . 10
1.2.4 Net transfer capacity of cross border flows . . . . . . . . . 11
1.2.5 Controllability of wind and solar power units . . . . . . . 12
1.3 Literature survey on the integration of high-shares of renewable
energies in Europe . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Priority infeed legislation: advantages and drawbacks . . . . . . . 18
1.5 The Power Node modeling framework . . . . . . . . . . . . . . . 20
1.6 MPC application in power system dispatch . . . . . . . . . . . . . 23
1.7 Upgrading the future EU power system . . . . . . . . . . . . . . 24
I Modeling Framework 26
2 Modeling data 28
2.1 IRENE-40 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.1.1 Load forecasts . . . . . . . . . . . . . . . . . . . . . . . 30
2.1.2 Wind energy . . . . . . . . . . . . . . . . . . . . . . . . 30
2.1.3 Solar energy . . . . . . . . . . . . . . . . . . . . . . . . 31
2.1.4 Net transfer capacities . . . . . . . . . . . . . . . . . . . 32
2.1.5 EU generation mix . . . . . . . . . . . . . . . . . . . . . 32
2.2 EURELECTRIC . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.1 Detailed hydro capacity data . . . . . . . . . . . . . . . . 33
2.2.2 Equivalent ramp rates of a pool of conventional generators 34
viii
CONTENTS ix
2.3 Swiss Federal Office of Energy . . . . . . . . . . . . . . . . . . . 35

2.3.1 Run-of-river and hydro reservoir hourly power infeed . . . 35
2.3.2 Simulating hydro infeed in rest of EU . . . . . . . . . . . 35
2.4 IEA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.4.1 Marginal cost of production of power plants . . . . . . . . 36
2.4.2 Cost projections for 2020 . . . . . . . . . . . . . . . . . . 37
2.5 Final structure of one country bus . . . . . . . . . . . . . . . . . 38
3 Modeling power nodes 40

3.1 Conventional load . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2 Generation with curtailable supply . . . . . . . . . . . . . . . . . 42
3.2.1 Non-buffered renewable plants . . . . . . . . . . . . . . . 42
3.2.2 Buffered hydro reservoir . . . . . . . . . . . . . . . . . . 43
3.3 Generation with controllable supply . . . . . . . . . . . . . . . . 44
3.3.1 Non-buffered conventional plants . . . . . . . . . . . . . 44
3.4 Storage technologies . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4.1 Pump storage hydro plant . . . . . . . . . . . . . . . . . 45
3.5 An interconnected European power node system . . . . . . . . . . 46
4 The dispatch simulator 49

4.1 Structure of the simulation tool . . . . . . . . . . . . . . . . . . . 50
4.2 Discretization of the power node system . . . . . . . . . . . . . . 51
4.3 Economic dispatch optimization . . . . . . . . . . . . . . . . . . 52
4.3.1 Characteristics of the optimization strategy . . . . . . . . 53
4.3.2 MPC controller parameters . . . . . . . . . . . . . . . . . 54
4.4 Controller behavior . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.5 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . 60
II Results and discussion 62
5 Simulation vs Real-World dispatch results 64

5.1 Comparison of energy mixes . . . . . . . . . . . . . . . . . . . . 65
5.1.1 Austria . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.1.2 France . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.1.3 Germany . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.1.4 Poland . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.1.5 Sweden . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.1.6 Switzerland . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2 CO2 emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6 The impacts of a high-share of renewable energy supply on the Euro-

pean power system 71
6.1 The impact of renewable policies . . . . . . . . . . . . . . . . . . 72
CONTENTS x
6.1.1 Scope of analysis . . . . . . . . . . . . . . . . . . . . . . 72

6.1.2 Dispatch results . . . . . . . . . . . . . . . . . . . . . . . 73
6.1.3 Renewable energy integration . . . . . . . . . . . . . . . 75
6.2 The impact on power exchange patterns . . . . . . . . . . . . . . 76
6.3 The impact on power markets . . . . . . . . . . . . . . . . . . . . 79
6.3.1 The electricity market price forecast . . . . . . . . . . . . 79
6.3.2 The electricity generation cost . . . . . . . . . . . . . . . 80
7 Sensitivity analysis results 83

7.1 EU level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
7.1.1 RES integration results . . . . . . . . . . . . . . . . . . . 84
7.1.2 Synthesis graph . . . . . . . . . . . . . . . . . . . . . . . 87
7.1.3 Value of Flexibility in the European power system . . . . 89
7.2 Country level: Germany & Spain . . . . . . . . . . . . . . . . . . 90
7.2.1 RES integration results . . . . . . . . . . . . . . . . . . . 90
7.2.2 Flexibility value in Germany and Spain . . . . . . . . . . 92
7.3 CO2 emissions analysis . . . . . . . . . . . . . . . . . . . . . . . 94
III Conclusions and Recommendations 97

8.1 The impacts of a high variable renewable supply . . . . . . . . . . 99
8.2 The value of flexibility . . . . . . . . . . . . . . . . . . . . . . . 100
8.3 Policy recommendations . . . . . . . . . . . . . . . . . . . . . . 101
8.4 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
A MPC power system dispatch 103

A.1 Economic dispatch where renewable plants compete with conven-
tional energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
A.2 Influence of quadratic ramping cost on the power system dispatch 105
B Impact of a high-share renewable energy supply 108

B.1 Power exchanges in Northern Europe . . . . . . . . . . . . . . . 108
B.2 Power exchanges in Central Europe . . . . . . . . . . . . . . . . 109
B.3 Power exchanges in Eastern Europe . . . . . . . . . . . . . . . . 109
B.4 Averaged spot market price in each European country . . . . . . . 110
B.5 Averaged generation cost of electricity in each European country . 110
C Sensitivity analysis results 111

C.1 Technology specific RES integration results in Spain . . . . . . . 111
D Data reference 112

D.1 Capital investment cost for power lines planning in Europe [1] . . 112
D.2 ENTSO-E net transfer capacities matrix . . . . . . . . . . . . . . 112
Bibliography 115
List of Abbreviations
TSO Transmission system operator

EPS European power system
RES Renewable energy supply
var-RES Variable renewable energy supply
VRP Variable renewable plant
VRE Variable renewable energy
IRENE Infrastructure roadmap of energy networks in Europe
IEA International energy agency
EURELECTRIC European Union of the Electricity Industry
ENTSO-E European network of transmission system operators for electricity
CCGTPP Combined cycle gas turbine power plant
NTC Net transfer capacity
MPC Model predictive control
PV Photovoltaic
CSP Concentrated solar power
PSP Pump storage plant
PSL Power system laboratory
O&M Operation and maintenance
xi
List of Figures
1.1 Installed PV Power Capacity in the European Union at the end of

2011 [2]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Installed Wind Power Capacity in the European Union at the end
of 2011 [3]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Comparison of the PV current trend in EU with the NREAP [2]. . 6
1.4 Comparison of the wind power current trend in EU with the NREAP
[3]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Coordinated control of TSOs set-points and wind farm output [4]. 12
1.6 Existing operational flexibility in power systems [5]. . . . . . . . 13
1.7 Flexibility Assessment Method of the IEA [6]. . . . . . . . . . . . 13
1.8 VRE deployment potential today from a balancing perspective [6]. 14
1.9 Flexiblity of conventional power generation technologies [7]. . . . 15
1.10 Evolution of generation in Spain on March 3 2010 [7]. . . . . . . 15
1.11 Total amount of electricity (logarithmic scale) that can currently be
stored in one ideal pumping cycle [8] . . . . . . . . . . . . . . . . 16
1.12 Pump storage function of Austria, France, United Kingdom, Spain
and Switzerland in the first 90 hours of an ideal pumping cycle [8] 16
1.13 Developed and remaining technically feasible hydropower poten-
tial in Europe per country in TWh [8] . . . . . . . . . . . . . . . 17
1.14 Electricity price in the German power exchange in 2007 and 2011
[9]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.15 Yearly MW installed power generating capacity in the EU [10]. . 19
1.16 Notation of a Single Power Node [11] . . . . . . . . . . . . . . . 21
1.17 MPC application in power system dispatch. . . . . . . . . . . . . 24
2.1 Final electricity demand for the EU in the five scenarios in TWh
per year [12]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Monthly consumption (January) as a percentage of the annual con-
sumption for Italy - historical data, extrapolated data, trends [12]. . 30
2.3 Measured historical wind power production in western Denmark
for third week in 2000 compared to simulated data, using Re-analysis
wind data [13]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
xii
LIST OF FIGURES xiii
2.4 Simulated wind onshore and offshore power infeed on the third
week of January in Germany for different reference years [12]. . . 31
2.5 Simulated PV and CSP infeed on the first week of January in Spain
for different reference years [12]. . . . . . . . . . . . . . . . . . . 32
2.6 IRENE-40 generation mix in 2010 of Austria, Belgium, Bulgaria
and Czech Republic [12]. . . . . . . . . . . . . . . . . . . . . . . 32
2.7 IRENE-40 RES scenario: EU generation mix evolution from 2010
to 2050 [12]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.8 Run-of-river and hydro reservoir hourly infeed in 2010 in Switzer-
land. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.9 Country fleet structure according to controllability, ramping capa-
bility, efficiency and cost of power plants. . . . . . . . . . . . . . 39
3.1 External demand process ξdrv (t) representing the aggregated elec-
tricity demand in France in 2010. . . . . . . . . . . . . . . . . . . 41
3.2 System view of interconnected power nodes representing the west-
ern European power system. . . . . . . . . . . . . . . . . . . . . 47
4.1 Basic structure of the MATLAB-based simulation tool. . . . . . . 50

4.2 Priority dispatch of the German power system in 2020 under a
high-share renewable infeed during the month of March. . . . . . 57
4.3 Economic optimum dispatch of the German power system in 2020
under a high-share renewable infeed during the month of March. . 58
4.4 Structure of the MATLAB-based simulation tool with the imple-
mentation of a sensitivity analysis. . . . . . . . . . . . . . . . . . 60
5.1 Power node dispatch in Austria vs IEA references in 2010 [14]. . . 65

5.2 Power node dispatch in France vs IEA references in 2010 [14]. . . 65
5.3 Power node dispatch in Germany vs IEA references in 2010 [14]. . 65
5.4 Power node dispatch in Poland vs IEA references in 2010 [14]. . . 65
5.5 Power node dispatch in Sweden vs IEA references in 2010 [14]. . 65
5.6 Power node dispatch in Switzerland vs IEA references in 2010 [14]. 66
5.7 CO2 emissions comparison: Power Node vs IEA reference. . . . . 69
6.1 Influence of renewable policies on renewable integration results. . 76

6.2 Net balancing exchanges between France and its neighbors in 2010
according to the ENTSO-E values [15]. . . . . . . . . . . . . . . 77
6.3 Line loading in 2020 around France in a low-share renewable en-
ergy supply scenario. . . . . . . . . . . . . . . . . . . . . . . . . 77
6.4 Line loading in 2020 around France in a high-share renewable en-
ergy supply scenario. . . . . . . . . . . . . . . . . . . . . . . . . 78
6.5 Forecast of the market price of electricity in Portugal. . . . . . . . 79
6.6 Forecast of the electricity generation cost and market price in Eu-
rope when the O&M cost of VRPs is considered. . . . . . . . . . 80
LIST OF FIGURES xiv
7.1 RES integration sensitivity relative to the power lines NTC and
storage power rating expansions in Europe. . . . . . . . . . . . . 84
7.2 RES integration sensitivity relative to storage power and storage
energy expansions in Europe. . . . . . . . . . . . . . . . . . . . . 85
7.3 RES integration vs increase of each flexibility parameter and their
combination. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7.4 RES integration sensitivity relative to power lines NTC and storage
power expansions in Spain and Germany. . . . . . . . . . . . . . 90
7.5 RES integration sensitivity relative to storage power and storage
energy expansions in Spain and Germany. . . . . . . . . . . . . . 92
7.6 CO2 emissions of the European power sector under different power
system configurations. . . . . . . . . . . . . . . . . . . . . . . . 94
A.1 Economic optimum dispatch of the German power system in 2020

under a high-share renewable scenario in which VRE plants com-
pete with conventional technologies. . . . . . . . . . . . . . . . . 104
A.2 German TSO keeps 75% of the fleet turned on and the quadratic
cost constraint is not changed. . . . . . . . . . . . . . . . . . . . 106
A.3 German TSO keeps 75% of the fleet turned on and the quadratic
cost constraint is removed. . . . . . . . . . . . . . . . . . . . . . 107
B.1 Northern European line loading in a low-share renewable supply. . 108

B.3 Central European line loading in a low-share renewable supply. . . 109
B.4 Central European line loading in a high-share renewable supply. . 109
B.5 Eastern European line loading in a low-share renewable supply. . . 109
B.6 Eastern European line loading in a high-share renewable supply. . 109
B.2 Northern European line loading in a high-share renewable supply. 109
B.7 Forecast of the spot market price in Europe with increasing RES
fixing the power system configuration to 2020 and the load demand
to 2050. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
B.8 Forecast of the electricity generation cost in Europe with increasing
RES fixing the power system configuration to 2020 and the load
demand to 2050. . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
C.1 PV and CSP energy integration sensitivity relative to power lines

NTC and storage power expansions in Spain. . . . . . . . . . . . 111
C.2 Wind onshore and offshore energy integration sensitivity relative
to power lines NTC and storage power expansions in Spain. . . . . 111
D.1 Selected interconnector projects in northern Europe [1]. . . . . . . 112

D.2 Selected interconnector projects in northern Europe [1]. . . . . . . 112
D.3 Indicative ENTSO-E net transfer capacities in Summer 2010. . . . 113
D.4 Indicative ENTSO-E net transfer capacities in Winter 2009-2010. . 114
List of Tables
1.1 Characteristics of exemplary pump hydro plants in Europe [16] . . 9

1.2 Notation of a Single Power Node [11] . . . . . . . . . . . . . . . 22
2.1 Overview of the five IRENE-40 demand and generation scenarios

[12]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Start up time and ramping constraints of conventional power plants
[7]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3 Equivalent ramp rates of a pool of aggregated conventional gener-
ators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4 Switzerland hydro mix yearly production [17]. . . . . . . . . . . . 35
2.5 Marginal Cost of Production of conventional power plants [18, 19]. 36
2.6 Marginal Cost of Production of renewable power plants [18]. . . . 37
2.7 Forecast of the marginal cost of power plants in 2020. . . . . . . . 38
3.1 Characteristics of the Power node i = 1, which represents the ag-

gregated load demand in France. . . . . . . . . . . . . . . . . . . 42
3.2 Non-buffered renewable Power Node numbering list. . . . . . . . 42
gregated solar PV installations in any EU country. . . . . . . . . . 43
gregated hydro reservoirs installations in Switzerland. . . . . . . . 44
3.5 Non-buffer conventional Power Node numbering list. . . . . . . . 44
gregated single gas turbine power plants installed in any EU country. 45
gregated pumped storage plants installed in Switzerland. . . . . . 46
5.1 Power node dispatch vs IEA reference values in 2010. . . . . . . . 68

5.2 Power node dispatch vs IEA reference values in 2010. . . . . . . . 68
6.1 Comparison of dispatch policies. . . . . . . . . . . . . . . . . . . 73

6.2 Priority infeed vs economic dispatch. . . . . . . . . . . . . . . . . 74
6.3 The influence of policies on the integration of each type of renew-
able energy infeed. . . . . . . . . . . . . . . . . . . . . . . . . . 75
xv
LIST OF TABLES xvi
6.4 NTC installed in the European power system of 2020 . . . . . . . 77
7.1 Operation cost of system update (bn € per year). . . . . . . . . . 85

7.2 Operation cost of system update (bn € per year). . . . . . . . . . 86
7.3 Flexibility measure for the ramping rates of conventional generators. 87
7.4 The value of flexibility (€bn per year) of the European power system. 89
7.5 Capacity forecast (in GW) in 2050 of solar and wind power plants
in Germany and Spain [20]. . . . . . . . . . . . . . . . . . . . . . 90
7.6 The value of flexibility (€bn per year) in Germany and Spain. . . 93
Chapter 1
Introduction
Since the Kyoto protocol was established in 1990, Europe has shown deter-
mination in fighting climate change and is leading the world in doing so. All
EU-members have recently renewed their commitment to the protocol at the Doha
climate change conference and are binded to reduce their carbon emissions by 18%
below their 1990 levels until 2020 [21]. Today, the EU is on track and has reached
an estimate of 16% below its 1990 levels [22]. Furthermore, the European Com-
mission has presented an ambitious roadmap for moving to a low-carbon economy
in 2050 which targets an 80% emission reduction [23]. To reach such reductions,
the first lever is the power generation sector, which holds the greatest emission
reduction potential and is thus targeted to be almost fully "decarbonised", reach-
ing 95% emission reduction in 2050 [23]. Hence EU members acknowledge that
continuing to burn solely fossil fuels as a mean to generate electricity is no longer
viable to meet the demand. The gradual switch to renewable energy is imminent
and necessary to first, sustain the EU’s development on a long term and second, to
increase its energy security and hedge itself against increasing volatile fossil fuel
prices.
To provide a reliable and secure electricity supply, power system operators
have to balance more or less instantaneously the production and consumption of
electricity, because it cannot be stored on a large scale. Traditionally, conven-
tional power plants were dispatched according to foreseeable demand fluctuations
in form of yearly (summer vs winter), weekly (weekday vs weekend) and daily
(day vs night) variations [7]. Solar and wind energy, being weather dependent, add
a variability and uncertainty to the supply of electricity and, thus, challenge the
task of balancing the system. An imbalance between supply and demand will re-
sult in a lower quality of the delivered electrical energy [24], which affects in turn
the proper functioning of any equipment connected to the grid. In the worst case,
a complete power system collapse, known as "black-out", can happen if this im-
balance is not regulated. At low wind and solar shares, the philosophy of "connect
and manage" has so far worked with transmission operators in balancing the system
[6]. When the wind blows or the sun shines, the energy generated from these VRP
2
CHAPTER 1. INTRODUCTION 3
is fed directly to the grid and conventional power plants are dispatched according
to the residual demand, which is the load demand minus the renewable energy that
was fed in. However, since Europe is counting predominantly on wind and solar
energy to decarbonise its power sector, the fluctuations in the residual demand will
be significantly higher, more frequent and sometimes unpredictable, leading to in-
creasing risk of a blackout. The existing operational flexibility in power systems is
vital in preventing such event. It emanates from five existing sources: energy stor-
age technologies, ramping capability of generators, power exchanges between in-
terconnected countries, demand-side participation and controllability of wind and
solar units. Once the role of each flexibility source in integrating high shares of
VRE is assessed and the overall limit of the system is found, the European power
system can be upgraded accordingly to increase this limit and stay secure under a
high share of VRE. Furthermore, given that a high-share of RES will have multiple
different impacts on the EPS, it would be interesting to quantify the operational
benefits when moving from low to high RES-shares and study the influence dif-
ferent renewable policies could have on these benefits. As well, it is intriguing to
investigate how will the electricity generation cost evolve and how will cross bor-
der power flows in the EU change in such a scenario. Most importantly will the
CO2 emission reduction target be achieved with only more wind and solar energy
or is a radical change in phasing out polluting plants also needed ?
This master thesis provides answers to the mentioned topics. It has the follow-
ing structure: Chapter 1 starts by presenting the hard facts behind our motivation.
We clarify the characteristics of each flexibility source and present a literature sur-
vey on assessing their roles for high-renewable penetration in Europe. We identify
research gaps and stress specifically the inefficiency of giving priority infeed to
renewable energies in a high RES-share scenario. We introduce the Power Node
Modeling framework and the algorithm of our simulator to tackle those gaps and
allow for a true system optimization. We explain our approach in finding the opti-
mal way to upgrade the European power system for a high renewable supply. Part
I consists of Chapter 2-4, which present our modeling framework: the data we
used and restructured from various sources (IRENE 40, EurElectric and IEA), our
EU interconnected power node model and our MATLAB based dispatch simula-
tor. Part II consist of Chapter 5, 6 and 7, which presents our findings. Chapter 5
features the validity of our simulation results for the year 2010 by comparing the
obtained theoretical dispatch in selected countries to literature. Chapter 6 analyzes
the impact that high share of RES will have on power flow exchanges and on the
generation cost and market price of electricity in Europe. We discuss also how dif-
ferent renewable policies affect the system operation cost in a high-VRE scenario.
Chapter 7 represents our core results in finding out how to develop the EPS. We
discuss the sensitivity of VRE integration to the increased capacity of each flexi-
bility source and value them, on a EU level and a chosen country level (Germany
and Spain). We present a solution to decarbonize the power system. Chapter 8 re-
capitulates Part II and highlights the main messages we have derived to minimize
the operation cost of the EPS under high-share RES, while maximizing the RES
integration and minimizing CO2 emissions. Policy recommendations are listed.
Future work in the field is recommended.
1.1 Motivation
The motivation of the thesis originates from the fact that Europe is leading the
world today in renewable energy investments and will start to face serious balanc-
ing challenges due to the variability and uncertainty of solar and wind energy. In
2011, the cumulative installed capacity of wind power plants in Europe (includ-
ing offshore Wind) has reached 94 GW [3], and the cumulative installed capacity
of solar power plants (only photovoltaic) has reached 51 GW [2]. Europe’s wind
power capacity surpasses the one of China and United States by 20 and 35 GW, re-
spectively, [25] and the solar power capacity surpasses by 45 GW the one installed
in each country [26, 25]. Figures 1.1- 1.2 illustrate the solar and wind installed
capacities in the EU by the end of 2011.
Figure 1.1: Installed PV Power Capacity in the European Union at the end of 2011
[2].
Figure 1.2: Installed Wind Power Capacity in the European Union at the end of
2011 [3].
Germany appears to be the leading EU country with more than 29 GW of Wind

and 24 GW of PV installed by the end of 2011. This represents almost a third of
the total capacity of power plants installed in Germany [27] and covers 12% of
yearly electricity demand in Germany. This significant capacity leads eventually to
very high variable infeed of solar and wind energy when the sun is shining or the
wind is blowing. On May 26 2012, at peak sun hour, Germany satisfied 50% of its
electricity demand with solar energy [28]. As for wind production, January 2012
was a record breaking month in Germany in which it covered 18% of the national
demand [29]. More astonishing, in Denmark, on March 18 2013 at 00:04, wind
energy satisfied 100% of the electricity demand [30], a breakthrough in renewable
energy generation. Higher renewable infeed are forecast to come since the EU is
following its National Renewable Energy Action Plan (NREAP) in 2020 [2, 3].
As depicted in figure 1.4 and in figure 1.3, the current trend of installing wind
power plants is well in line with the NREAP and the trend of installing PV plants
is surpassing the NREAP by 1.5 times.
Figure 1.3: Comparison of the PV current trend in EU with the NREAP [2].
Figure 1.4: Comparison of the wind power current trend in EU with the NREAP
[3].
Assuming minimum standard load factor for PV (0.1) and for wind energy
(0.2) [6], this will lead the EU to satisfy at least 17% of its actual electricity de-
mand with VRE in 2020 and 45% in 2050 if the trend continues at a linear pace.
With increasingly volatile supply, if the sun does not shines nor the wind blows, the
loss of load probability will increase considerably, putting the economy and wealth
of EU at high risk of a blackout. To overcome this challenge, power flow exchanges
between interconnected countries, energy storage, ramping of conventional gener-
ators, demand side management and curtailment of wind and solar power units are
valuable sources that provide balancing services to keep the system secure. Hence,
it is a must to evaluate the limit of these flexibility options in integrating VRE with-
out risking the reliability of the European power system . Once this limit is found,
the question is how to optimize it, thereby motivating our work. We explain now
the characteristics of each source and their respective operational limits.
1.2 Sources of operational flexibility in the grid

Power system management is not only a matter of scientific and technical know-
how. Its complexity drives it to be "something of an art" [6]. A grid operator
disposes of many tools to fine tune the balance between supply and demand. Each
tool has a specific role, a specific cost and operates under specific technical con-
straints in balancing the power. To minimize the balancing cost, understanding the
flexibility properties of each source is required.
1.2.1 Ramping capability of generators

The operation of all fossil power generators (nuclear, coal, oil and gas) is similar.
They burn fuel in a combustion chamber, extract its thermal energy and convert
it into mechanical power via a turbine. The turbine feeds an electric generator
which converts the rotational power into electricity 1 . Due to the nature of the
fuel and the type of turbine used, these plants differ in the time delays between
each thermodynamic process. Nuclear and coal power plants are equipped with
steam turbines while gas and oil power plants use gas turbines. Steam turbines
expand high pressure steam to create mechanical power, while gas turbines directly
expand pressurized flowing gases. Since gas turbines do not need any pre-heated
steam to operate, they can be quickly brought online. As for steam generation,
the fission process in a nuclear reactor is a slow dynamic process compared to
burning fuel in a combustion chamber, therefore bringing a nuclear plant from
start to nominal capacity takes up to three days. All these type specific ramping
and start-up constraints are crucial for evaluating the operational flexibility of each
conventional plant in balancing variable renewable output. The literature survey
provides single ramping constraints data for each plant and we explain in Chapter
2 how we adapted single plant constraints to a pool of aggregated plants per country
and how we derived the cost production function of each.
Regarding the ramping capabilities of biomass and geothermal plants, since
they are equipped with steam turbines, they are considered to have the same ramp-
ing capability of coal plants. For hydro plants, the water inflow in the turbine
blades defines the plant output. Since this water flow provides mechanical power
that can be translated into electrical power very fast (be it in the penstock of a hy-
dro reservoir or the natural flow of a run of river plant), hydro plants offer the best
1
A generator follows the principle of Faraday’s law of Induction, which states that a rotating coil
placed in a varying magnetic a field will produce an electric current [31]
operational flexibility and can be ramped up to full capacity in less than 15 minutes
[8].
1.2.2 Energy storage

Till today, no solution has been found for bulk electricity storage. Electricity has
to be converted into an intermediate storable form of energy, which in turn can be
converted by inverting the process. The common intermediate conversion forms of
energy are chemical, mechanical and thermal energy. The common technologies
that provide such conversions on a large scale are batteries, pump hydro storage
(PHS) plants and compressed air energy storage (CAES) plants.
Batteries
Electricity is stored in batteries by inversing the reaction that allows the gal-
vanic cell to produce a current. The cell consists of two electrodes, the anode which
consists of special compounds that permit an electron to flow out of it, the cathode
which consists of compounds that receive the electrons. When inserted in a ionic
electrolyte solution and connected via a circuit, a chemical reaction takes place and
produce electricity until the electrodes run out of their compounds. The battery is
here fully discharged. Storing electricity forces the inverse chemical reaction until
the cell is fully recharged. Herein, the size of a battery determines its capacity and
its cycling time, thus the flexibility it can offer to the grid operators. Large batteries
that can provide balancing services on the MW scales are seldomly deployed due
to their high capital cost. The largest battery in the world was deployed last year
in the province of Hebei in China, "taking up the area of a football field", with
a capacity of 36 MW dischargeable in one hour [32]. Its cost was evaluated to
$500 million. As it stands today, the potential of batteries in providing profitable
flexibility services in the form of large scale energy storage does not lie in over-
sized battery plants [33]. The true potential lies in Electric Vehicles (EV) which
represent a significant decentralized small-scale storage source that can be smartly
called upon to either, absorb excess power in case of a higher than predicted VRE
infeed, or, inject power in case of a lower than predicted VRE infeed [34]. If many
EVs are given the right economic incentive to charge or discharge their battery, the
aggregation of a large pool of small scale storages becomes a significant source of
operational flexibility. The project "IRENE-40" unveils in its deliverable [35] that
an optimized EV charging strategy, in a 30% wind share scenario, helps reduce the
annual operating cost of the UK power system by 2% when EV have penetrated a
quarter of the automotive market. These cost savings comes to £230 million a year
and are attributed to minimizing wind energy curtailments and maximizing the use
of least-expensive conventional plants. With 34.6 million of vehicles licensed in
the UK [36], the power system will be saving on operation cost £27 per EV in the
UK.
Pump hydro storage

PHS plants are the oldest and most mature technology used in storing electric-
ity in the form of mechanical energy. The first plant was built in 1892 in Zürich,
Switzerland [37]. The surplus of electricity powers a pump to elevate a volume
of water to an upper reservoir at a predefined height. When electricity is needed,
water is run downwards and its potential energy is converted to electricity. The
flow rate of water turns the blades of a turbine, the turbine feeds a generator which
converts the rotational energy to electricity. Thus, the energy storage capacity of
PHS plants depends on the volume of the reservoirs and the elevation difference
between them. The power generation capacity depends on the power rating of the
pump and turbine installed. We constructed a list of pump hydro plants in some
European countries that illustrates these characteristics.
Table 1.1: Characteristics of exemplary pump hydro plants in Europe [16]
Location Reservoir (Mm3 ) Head (m) Capacity (GW) Energy (GWd)

IT 27 1048 1.3 0.44
FR 140 955 1.8 15.4
GR 6.3 288 1.1 0.21
LU 6.8 290 1.7 0.21
A sample calculation for the PHS plant Grand’Maison in France is shown2 .

The list shows that the energy capacity of pump storage plants is by 1000 times
bigger than the largest battery installed but it is very much dependent on a coun-
try’s geographical assets. The great advantage of PHS though is their almost in-
stantaneous dispatchability since water will flow very fast from the upper to lower
reservoir and the conversion from rotational to electrical energy is almost instanta-
neous [24]. All plants can therefore be ramped up or down at full capacity in less
than a minute. Thus in France, if the Grand’Maison upper reservoir is full, the plant
is able to provide 370 GWh of backup capacity before its runs out. The Vianden
PHS plant in Luxembourg could store 5 GWh of excess renewable energy. One
clearly recognizes the imminent role PHS plays in providing operational flexibility
to the grid. Please note, the energy value calculated in Table 1.1 represents the ideal
amount of electricity that can be stored in one day if there were no losses in the
system. Depending on the age of the plant, the efficiency of pump storage ranges
between 70 to 85% [8] making them the best available energy storage technology
today.
2
Eideal = m.g.h = ρ.V.g.h = 998 mkg3 ∗9.79 sm2 ∗140.106 m3 ∗955m∗ 3.6
1 kWh 1
MJ 106
GWh 1 h
∗
kWh 24 day
= 15.4GWd
Compressed air energy storage

The first CAES plant was patented by Stal Laval in 1949 [38]. The concept
behind it is to use electricity to compress air and store it in an underground reser-
voir. The reservoir consists of salt cavern cavities which reaches 700-900m below
ground and conserve the high pressure of the air [39]. Once electricity is needed,
the pressurized air is extracted from the cavern and expanded in a gas turbine,
which feeds a generator. To increase the efficiency of the system, air is cooled
down during compression via inter-coolers and heated up before its expansion by
burning gas inside the turbine. There exists only one CAES plant in Europe. It
was commissioned in 1978 in Huntorf in Germany to offer black-start2 capabilities
to nuclear units near the North Sea [38] and to provide inexpensive peak power
[38]. The volume of the cavern is 300 000 m3 and can store or deliver up to
580 MWh of electricity in 2 hours of operation with an efficiency of 42% [39].
Since CAES technology is closely related to gas turbine technologies, they can
be brought online at full capacity in less than 2 hours [7] and allow very flexible
operation. However, despite their technological success, CAES have not been de-
ployed during the 80’s and 90’s due to a lack of economic incentives to increase
efficiency and decrease capital cost [41]. Engineers have thus been motivated to
find a solution on their own. The promising technological development of CAES
is the advanced-adiabatic CAES (AA-CAES) which consists of capturing and stor-
ing the heat of the compression stage, to later use it in re-heating the air during the
expansion stage. The technology would become carbon-free since the combustion
of gas is no longer needed and more profitable since it will be cutting on fuel and
carbon emission cost. Moreover, it should be possible to bring them online in less
than 15 minutes offering better flexibility [39]. This technology is forecast to be
deployed before 2020.
1.2.3 Demand side participation

Demand side participation can be categorized in two parts: demand side man-
agement and demand response. Demand response has triggered the need for the
traditional grid today to be "smart", as the "supply follows load" paradigm is shift-
ing towards a "load follows supply" paradigm [42]. This shift means that loads,
which were traditionally considered inflexible to the price of electricity and un-
controllable, offer a new degree of freedom for grid operators in balancing supply
and demand [43]. The smart grid platform allows for bi-directional energy flows
with bi-directional communications signals between producers and consumers of
electricity [43]. Therefore, when the supply falls short of demand and the elec-
tricity price increases, a consumer that is aware of this price increase will shift its
consumption to lower electricity price hours, thereby reducing the need to ramp up
an expensive power plant or discharging a storage plant. This shift in consumption
2
A black start is the process of restoring a power station to operation without relying on the
external electric power transmission network [40]
can be either manually done or automatically controlled via innovative controller

boxes [42]. This box controls the flexible load appliances of the house which can
store heat; heat pumps, electric heaters, refrigerators, air conditioners. These loads
will therefore always be controlled to consume electricity at low supply times,
store heat and be dispatched on consumer demand. MVV-energie, a regional util-
ity in Germany, conducted a practical test in 1000 houses in Mannheim, to evaluate
the potential of demand response (DR). By installing an "Energie-Buttler" (energy
buttler) in every house, they reach to the conclusion after one year of measurement
test, that the feasible potential of DR in Mannheim is 4% of a peak load of 420
MW. They were able to shift the peak for a duration of 30 minutes [44].
Demand side management are measures of last resort taken by electric utilities
to level out the balance by smoothing out peaks in demand to increase the efficiency
of the energy system [42].
1.2.4 Net transfer capacity of cross border flows

One of the vital tools of any market in balancing supply and demand is to import
the required supply in case of a shortage or export any surplus. To do so in the
European electricity market, transmission system operators (TSOs) of each EU
country are legally bound to assess the cross-border transmission capacity limit so
that the exporting or importing flows of one country do not compromise the security
of the whole network [45]. They start by exchanging a "base case" input data
set which relies on historical recorded load and generation measurements in each
country. They simulate cross-border flows between two areas according to the base
case while using an AC load flow model that includes all thermal ratings of network
elements as well as voltage and phase angle parameters [45]. Thermal ratings
are very much influenced by seasonality effects thereby ENTSO-E differentiates
between summer and winter cross-border flows. They further push the limit of
these cross-border flows by increasing the generation from the exporting country
and decreasing the generation from the importing country until a security limit is
breached [45]. The last value of the additional exchange that did not violate any
security constraint can be added to the base case cross border flows. This forms the
net transfer capacity (NTC) exchange between two countries which is a significant
asset for grid operators in balancing renewable energy. For example, Denmark with
the most installed wind capacity per 1000 inhabitants in 2011 reaching 706.2 kW
[3], is using extensively its power lines to integrate wind energy and satisfy 28%
of its annual electricity demand [46]. Energienet.dk states in [47] that Denmark
has imported and exported 11 TWh in 2008 which represented 30 % of its annual
consumption. Hereby, the contemporary Danish case exemplifies the importance
of evaluating accurately the net transfer capacities between countries to integrate a
high share fluctuating energy supply while keeping the European network secure.
1.2.5 Controllability of wind and solar power units

This last option of flexibility was not envisioned at the beginning of renewable
energies deployment. Till today, grid operators have been dispatching the power
system according to the residual load curve, which is derived from subtracting from
the load curve the predicted renewable output. They did so as renewable energies
are given priority infeed over any other type of generation. Renewable plants have
thus been considered as negative inflexible loads in traditional power system dis-
patching. Subsequently, if the night was very windy in Germany, the load balance
dropped below zero which translated in negative electricity prices in the power
market. This means that at this moment in time, nuclear power plants (NPP) were
forced to shut down their plants to keep the system frequency stable. To evade high
cycling cost, it was more beneficial for NPPs to pay to export electricity to neigh-
boring countries than to shut down and restart the plant when the wind stopped
blowing strongly the next day. This happened during five non-consecutive days
in January 2012 when the hourly German price decreased to -100.1 e/MWh [48].
This market inefficiency emphasized the need to develop means to control renew-
able power plant output. Today, power electronic industries have come up with
solutions to control the output of wind and solar power units, thereby no longer
constraining the grid operator to dispatch the system according to the residual load
curve. ABB has developed a Wind Turbine (WT) Controller box that allows TSOs
to dispatch the turbines to meet the requested set-points for balancing the power
system [4]. Figure 1.5 depicts the functioning of this box.
Figure 1.5: Coordinated control of TSOs set-points and wind farm output [4].
To provide control options according to grid codes requirement, the WF con-

troller aggregates the turbines output in one generation plant for the TSO and uses a
model predictive control algorithm to optimize the system. The reference dispatch
trajectory is set according to the set-points and is updated with wind farm weather
informations such that wind farm output is following the TSOs set-points. Such a
box can offer as well delta, frequency and voltage control thus promising to add a
great flexibility potential to grid operators.
SMA Solar Technology AG also came up with a innovative Power Control

Module that can be retrofitted to standard SMA inverters. The power control mod-
ule is able to remotely limit the active power from PV plants to 0%, 30%, 60%,
100% and can shut down a 6 kW plant within 50 ms [49].
We summarize in figure 1.6 the tools an operator disposes to fine tune the bal-
ance between supply and demand.
Figure 1.6: Existing operational flexibility in power systems [5].
The [+/-] sign indicate that each source can add or reduce power from the
balance. We present now an overview of previous work that dealt with evaluating
the role of these flexilibity sources in integrating high shares of VRE.
1.3 Literature survey on the integration of high-shares of

renewable energies in Europe
The IEA recent report on grid integration of renewables [6] is one of the many re-
ports we found that target this field of study. It assesses the ability of the European,
North American and Japanese power systems to balance the fluctuations of VRE
via the new Flexibility Assessment (FAST) method developed by the IEA. The ap-
proach in figure 1.7 is divided in four steps and builds on a series of key questions
that relates to the resources and nature of each power system area to identify the
potential of renewable energy penetration.
Figure 1.7: Flexibility Assessment Method of the IEA [6].
The first step consists of calculating the Technical Flexibility (TF) index of four
resources (dispatchable power plants, storage, interconnection resources and de-
mand side management) to ramp up and down over different balancing time frames
(15 minutes, 1 hour, 6 hours, 36 hours). This index is derived from summing the
MW output of each flexibility source at the respective timescale and normalizing
it to the peak demand of the area. Once found, IEA demonstrates that if the spe-
cific grid and market rule constraints of the power area are not included, this index
could be overestimated or underestimated for the given area, therefore providing a
false judgment on the ability to balance VRE. Hence, step two consist of adjusting
the Technical Flexibility index to the specific attributes of the power area in ques-
tion. In step three, the IEA determines the maximum flexibility requirement of
the system, by deriving the residual load (fluctuations of demand minus VRE) and
combining it to the contingencies of the system. The final step brings all elements
together to establish the Present VRE Penetration Potential (PVP), expressed as
percentage of gross electrical demand of the area in question. The main finding of
their research is illustrated in Fig 1.8.
Figure ES.2 VRE deployment potentials today from the balancing perspective
100%
90%
PVP (present VRE Penetration Potential
Height of bar shows deployment potential

80% based on technical flexible resource
70% 63%
Colour gradient highlights that flexible resource
60%
of gross electricity demand)
will not be fully available (see lower part of figure)

50% 48%
45%
40% 37%
31% 29%
30% 27%
19%
20%
10%
0%
Denmark Nordic United States NBSO area Great Britain Mexico Spain and Japan
market West (2017) (of Canada) and Ireland Portugal
Grid
Market
Score: High Medium Low
18 The availability of flexible resources for the balancing task is

constrained bydeployment
Figure 1.8: VRE a range ofpotential
power today
system attributes
from a balancing perspective [6].
The analysis has identified a range of characteristics, present to a greater or lesser extent in all cases
examined, which will constrain the availability of flexible resources to take part in balancing electrical
The Danish power system was the most apt to integrate renewable energy at a PVP
supply and demand. In some cases this will result in reduction of PVP to well below the values shown
ofin 63%
Figure from gross electricity
ES.2. Sub-optimal grid strength demand.
and market The Japanese
design power
are the most system
important was (although
of these the lowest
system
with operation
19%. Thetechniques
leap in PVP and many other resource-specific
is explained factors caninrepresent
by the difference serious barriers of
interconnectedness
also). The impacts of these two key characteristics in the case study areas are represented in the lower
the
parttwo systems
of Figure ES.2,with
with atheir
simpleneighbors
“traffic lights”andapproach.
the value Forgiven toreasons
practical flexibility
of data inavailability,
their power
it was not
market possible in this phase
configurations. of the analysis
Denmark to quantifyinterconnected
is strongly these constraints. Moreto itsrefined assessment
neighbors, with
using the FAST Method, with sufficient data availability, should attempt to quantify the impact of these
which it
constraints. constitutes one large balance area, whereas Japan is an isolated island and
constituting one balance area. If the line capacity played a major role in balancing
Weaknesses in parts of the existing transmission grid can cause the temporary separation of flexible
VRE in Denmark,
resources it was notand
from both demand-side theVRE
case for the rest
requirements of the areas,
for flexibility. where dispatchable
Such weaknesses may exist
for a number of reasons, but are particularly
generation capacity remained by far the most important source likely to be found at the borders of distinct balancing In
of flexibility.
areas within large power systems. Additionally, VRE power plants, particularly onshore wind, may be
fact, the Nordic market started with the highest TF index compared to other areas
located at a considerable distance from demand centres, where the wind resource is strong but where
(excluding Denmark)
the grid in contrast sinceweak.
is relatively it benefits
Reflecting from a large
strategic policycapacity
to deploy of VREhydro plants
capacity, (32∼
in-depth
studies to identify such grid weaknesses should be undertaken
GW) fully dispatchable in 15 minutes. With a peak demand of 69 GW and a flex- without delay. Remedial lead times may
be lengthy particularly if new transmission corridors are required and their rollout is likely to encounter
ibility index
opposition fromof 56%,
local the Nordic
communities. Measuresmarket
shouldprovided
be examined 38whereby
GW of operational
carrying capacity inflexibility
weaker
inareas
a 15 canmin
be augmented
timescale, through
84%advancedconsistinggrid technology
of hydroand operation
power. Thetechniques
study with negligiblethat
concludes
disruption to the local environment.
today’s power systems have greater capacity to handle fluctuating renewable elec-
Marketsthan
tricity should be (re)configured
commonly believed. so that the full flexible
However, resource
IEA states is able
that thetoflexibility
respond in resources
time to
© OECD/IEA - 2011
assist in balancing. Power markets should incorporate mechanisms that enable sufficient response
should still be increased for a better integration of renewable sources. Quantitative
from supply-side and demand-side flexibility assets. Electricity is usually traded through a combination
analysis
of long-term remains
bilateral to be done
contracts to find
and daily powerout monetary
exchanges. cost
Markets thatand benefits
rely heavily when
on the formermoving
will
into a high share RES scenario.
Harnessing variable renewables: a guide to the balancing challenge
In contrast to the previous study, the Union of the Electricity Industry - EUR-
ELECTRIC assessed solely in [7] the role of each flexible generation to balance
the deployment of VRE in Germany, Spain and the Nordic countries. In collabora-
tion with VGB PowerTech, EURELECTRIC surveyed conventional power plants
in Europe on matters of their flexibility, that is to say to the cold and warm start
up times of the specific plant, the rate of load following operations, the minimum
possible load and the necessary shut down time. "The figures come from genera-
tors’ daily business, technical values as claimed by equipment suppliers have not
been taken into consideration due to their lack of representativeness". The investi-
gated plants are nuclear power plants (’NPP’), hard coal-fired power plants (’HC’),
lignite-fired power plants(’Lign’), combined cycle gas-fired power plants (’CCG’)
and pumped storage power plants (’PS’). Their results is presented in figure 1.9.
Figure 1.9: Flexiblity of conventional power generation technologies [7].
Their finding proved that NPP were the most suited conventional power plants
to provide load following operations. In fact, 75% of their nominal output can
be ramped up or down in 15 minutes if they were already brought online. Thus,
constraining NPP to pure base load power operations diminishes the power system
flexibility potential in question. EURELECTRIC highlights that it is important
to differ between demand (load)-driven fluctuations and generation-driven fluctua-
tions. The load variations emanate from the traditional seasonal (summer vs win-
ter), weekly (working day vs week) and daily (night vs day) load demand changes
and the generation-driven variations can originate, for example, from the sudden
decrease in wind speeds or appearance of clouds, which in turn disturbs the power
system planning and requires quick power ramps. In that case, pumped storage
plants appear to be still the only valid technology that is able to cope with these
high power ramps and deliver active power on quarterly basis. However, if their
reservoirs get depleted, then CCGTPP plays a vital role in offering back up capac-
ity as they can reach full capacity from warm condition in two hours. Figure 1.10
illustrates the crucial role CCGTPP units played on March 3 2010 in Spain.
Figure 1.10: Evolution of generation in Spain on March 3 2010 [7].
At 4am on that day, the wind stopped blowing, reducing the output of wind plants
by 5 GW while the demand for electricity was going to increase by 10 GW over
the next 5 hours. The needed upward ramp in the grid was thus 15 GW in that
time frame to avoid a blackout. 27 CCG plants were turned on to offer such bal-
ancing. The report concludes, similarly to IEA [6], that the existing fleet of the
studied countries was flexible enough to provide balancing services for the next
years. Hydro is the backbone of the power system in providing flexiblity means
on short time scales and will play a crucial role for the integration of high share of
VRE. Thus any existing hydro potential in Europe should not be left unexploited,
a statement that motivated the organization to evaluate Europe’s Hydro potential.
In [8], EURELECTRIC provides facts and figures on hydropower in all EU
countries, detailing the power capacities of run-of-river plants, hydro reservoirs
and pump storage plants (PSP) and the energy rating of the latter. It also investi-
gates the unexploited hydro potential of Europe. According to a questionnaire they
distributed to pump storage plants in Europe, the organization was able to group
vital data on the energy capacity of these storage plants indicated in figure 1.11.
Data is missing for Italy, Romania and Sweden.
Figure 1.11: Total amount of electricity (logarithmic scale) that can currently be
stored in one ideal pumping cycle [8]
The value on the top of each bar represents the total amount of energy the country
can pump from all its lower reservoirs to upper reservoirs, at full capacity, until
the lower reservoirs are empty or the upper reservoirs are full (ideal pumping cycle
[8]). With a capacity of 2.8 GW of pump hydro plants in the United Kingdom
(UK), the country could offer 12 hours of full load operation to store an excess of
33 GWh of wind infeed or provide 33 GWh to balance a deficit in wind produc-
tion. However, EURELECTRIC stresses in their report that the duration of an ideal
pumping cyle differs from reality as not all power plants have the same reservoir
and power capacities. Figure 1.12 illustrate plant-specific data that was collected
from Austria, France, the UK, Spain and Switzerland.
The area under the red curve (UK) sums up to 33 GWh and spans 24 hours of op-
eration. It highlights the "temporal distribution" of pump storage plants in UK and
Figure 1.12: Pump storage function of Austria, France, United Kingdom, Spain
and Switzerland in the first 90 hours of an ideal pumping cycle [8]
falsifies our previous statement regarding the 12 hours of full load operation. In
real life, the UK pumped storage plants are only able to offer 5 hours of storage
operation at rate of 2.8 GW per hour. Then for 15 hours they could bid for 1.1
GW per hour to finally 2 hours at 0.6 GW and one hour at 0.4 GW, as their up-
per reservoir gets empty or full. Temporal distribution is a key characteristics to
for assessing the exact function of PSP in balancing variable renewable infeeds,
nonetheless, EURELECTRIC failed in obtaining plant-specific data from all EU
countries due to confidentiality reasons. For the purpose of our study, the power
and energy rating of each PSP in Europe was highly valuable. We will come back
to this point in our results section. Regarding the hydro potential, the report states
that half of the technical feasible potential is not exploited in Europe reaching 650∼
TWh per year. Country specific potential are presented in figure 1.13.
Figure 1.13: Developed and remaining technically feasible hydropower potential

in Europe per country in TWh [8]
The report concludes that pumped storage plants are highly suitable to take the role
of a large-scale battery for the integration of intermittent generation in Europe.
Categorizing the energy capacity of PSP per country is essential to evaluate the
true potential it can play in balancing variable renewables. Contrary to popular

belief, the total hydro potential of Europe has not been fully exploited but should be
since the technology is competitive, climate-friendly and contributes to the system
stability.
In [50] Greiner, Rasmussen and Andresen quantify the size of storage and bal-
ancing needs for the European power system under various future shares of wind
and solar energy. They derive hourly power generation time series for wind and
solar that cover the 27 European countries including offshore regions. They use
a "storage first" strategy where any deficit between supply and demand are first
backed by a storage reservoir unless it runs empty, and any excess is stored unless
the storage is full. Likewise, but with greater modeling accuracy than the previous
studies, the authors approve that no additional balancing is required in a perfect
European grid (no transmission constraint) if the average wind and solar power
production is ceiled at 50% of average load. To go above such shares, balancing
and storage needs will increase considerably, especially if the synergies between
them are not understood. The authors explain that a highly efficient 6-hour storage
can help integrate up to 75% of renewable energy without having more balancing
energy needs. The authors conclude that it is realizable to have a fully renew-
able scenario in Europe with an average wind and solar power infeed of 3% more
than the average load (av.l) on the condition that three services are procured: 150∼
TWh/year (5% of av.l) of balancing energy coming from hydro or biomass, a highly
efficient 6-h storage and 25 TWh of hydrogen storage for balancing seasonal vari-
ations in demand.
In similar studies which we have read [51][20][52], we found some missing
common points:
1. Taking into consideration the flexibility option of curtailing renewable plants

for balancing services by not prioritizing its infeed to the grid.
2. Developing a method that unifies traditional power system operation with

the availability and controllability of intermittent renewable energy sources
and energy storage.
3. Dispatching the power system with a prediction horizon and not on an hourly
basis to optimize effectively the cycling of storage plants.
4. Modeling the cross-border transmission constraints between EU countries.
5. Modeling the hourly infeed of all variable renewable supply in Europe; wind,
solar and hydro.
6. Modeling with accuracy the available storage size of pumped hydro in each
EU country (specific energy and power rating capacities).
7. Quantifying the value of each flexibility source in the integration of VRE.

8. Quantifying the operation cost of integrating high shares of renewable en-

ergy in the grid and the impact of different renewable energy policies.
Each listed point is implemented and presented in our work. The motivation
behind the first three points is successively introduced in this chapter. Point 4,5
and 6 are clarified in chapter 2. Point 7 and 8 will be found in the conclusion.
1.4 Priority infeed legislation: advantages and drawbacks

Europe would have not experienced such a renewable boom if it was not for the
right policies implemented at the right time. The liberalization of its power mar-
ket coupled with the emergence of the feed-in tariff scheme led to a propitious
investment climate. With the liberalization, the generation sector was privatized
allowing independent power producers to participate in the generation mix and in-
troducing competition for efficient electricity pricing. Independent power produc-
ers were mostly owners of small renewable plants as the capital investment in large
conventional power plants could not be afforded by small players. To encourage
the deployment of renewable technologies, governments legislated a law that gave
priority infeed to wind and solar power output over conventional generation. More-
over, governments offered a fixed long-term remuneration for this infeed, known as
"feed-in tariff", to ensure secure returns for investors in clean technologies. With
no surprise, it was first Germany that framed back in 2000 such policy, known as
the "German Renewable Energy Act". With the massive deployment of PV plants
in Germany feeding-in electricity with a zero marginal cost, the average electricity
price dropped in 2011 by 10% according to the German Institute for Future Energy
Systems [9]. Moreover, since PV output usually peak at noon when the demand of
electricity is highest, peak consumption prices were reduced by 40% in Germany
between 2007 and 2011 [9]. This observation is illustrated in figure 1.14.
Figure 1.14: Electricity price in the German power exchange in 2007 and 2011 [9].
By all means, with the success of the feed-in tariff policy in Germany, EU
countries started to adapt it progressively, leading to a steep increase in renewable
energy investments in Europe. The figure below exemplifies the continuous install-
ments of wind power plants and the boost in PV installations increase in EU after
2000, even after the financial crisis in 2008.
Figure 1.15: Yearly MW installed power generating capacity in the EU [10].

The coupling of the priority infeed and feed-in tariff policy has clearly proven
its effectiveness since renewable investments were not impeded by the financial
crisis. However, in a high-share renewable scenario, the priority infeed policy is
constraining power system operators in optimally dispatching the system. Com-
mon wisdom will dictate that the more available-free renewable energy is fed in
the grid, the more inexpensive the operation of the system will be. Nevertheless,
with more intermittent generation in the mix, the variability of supply imposed on
the system will increase and thereby, balancing cost likewise. There exists an op-
tion to reduce the need for balancing power by just simply curtailing renewable
output. Hence, if the reduction in balancing cost outweighs the benefit of priori-
tizing VRE, then the priority infeed policy is no longer economically optimal in a
high VRE scenario. Wu and Kapunscinski show in [53] that curtailing wind power
output helps reduce the operation cost of the Michigan power system by 8% when
the shares of wind energy have reached 34 % of national production. This signifi-
cant value comes from the fact that they have detailed the production cost of their
fleet, dividing it in intermediate and highly rampable fleet and differentiating be-
tween start up fuel and part loading cost. Hence, they could prove that the option of
using more efficiently the cheaper intermediate fleet at the extent of curtailing wind
energy was more economic than fully integrating it, thereby being constrained to
dispatch more often the expensive peaking unit in case of sudden shortage. This
study motivated us to question the economic optimality of the priority infeed pol-
icy in Europe under a high share scenario of VRE. Section 6 will elaborate on this
issue. It is important to note that the phase out of the priority infeed policy is not
correlated to the phase out of the feed-in tariff policy. The latter can still be present
even if VRE are no longer given priority infeed. We do not discuss in our work if
the feed-in tariff would still be justifiable with more renewables in the mix. This
financial assessment falls outside of the scope of study.
1.5 The Power Node modeling framework

Traditional power system modeling have not been designed in a unified framework
thats includes information on the observability and controllability of fluctuating
power system units, the energy constrained capacity of storage units and the flexi-
bility of demand units; it has been focused on the interactions between the electric
power injected into the grid or electric power demanded from the grid [5]. Kai
Heussen, Stephan Koch, Andreas Ulbig and Professor Göran Anderson, from the
Power Systems Laboratory of ETH Zürich, have developed a method that allows to
model power system units as such, thereby including all degrees of freedom in op-
erating the system to assess the value of flexibility of each power system unit. Their
work will be introduced in this section and is based on [5, 54, 11, 55]. Figure 1.16
illustrates the structure of a single power node which consists of a storage capacity
C normalized to 0 ≤ x ≥ 1 and which is embedded between the demand/supply
process domain and the grid domain.
Figure 1.16: Notation of a Single Power Node [11]
The idea behind the approach is that any power source or sink connected to the
electric grid requires the conversion of some form of energy into electricity with
a certain efficiency and vice versa. The demand and supply process are lumped
into parameter ξ, with ξ < 0 indicating use and ξ > 0 supply. In the grid domain,
ugen describes a conversion corresponding to power generation with efficiency ηgen ,
while uload describes a conversion corresponding to power consumption with effi-
ciency ηload . If this conversion happens to take place via a buffer with a certain
capacity C e.g a battery, then the external process ξ is decoupled from the grid re-
lated exchanges ugen and ηload . The storage losses of the buffer are modeled by the
term ν. The possibility to curtail an external supply/demand process is lumped in
the parameter w, with w > 0 indicating curtailment of provided energy and w < 0
curtailment of demanded energy. All these energy flows will influence the dynam-
ics of the Power Node storage level x. Mathematically expressed, the dynamics
of an arbitrary power node i ∈ N = {1, ..., N} will follow the following equation
subjected to a generic set of constraints.
Ci ẋi = ηload,i uload,i − η−1

gen,i ugen,i + ξi − wi − νi (1.1)
s.t (a) 0 ≤ ximin ≤ xi ≤ ximax ≤ 1
(b) 0 ≤ umin max
gen,i ≤ ugen,i ≤ ugen,i ≤ 1
(c) 0 ≤ umin max
load,i ≤ uload,i ≤ uload,i ≤ 1
(d) 0 ≤ ξi .wi
(e) 0 ≤ |ξi | − |wi |
(f) 0 ≤ νi ∀i = 1, ..., N
The constraints implies that in a) the state of charge is normalized and bounded,
(b, c) the grid-variables are non-negative and bounded, (d, e) the external sup-
ply/demand and the curtailment need to have the same sign and the latter can not
exceed the former, (f) storage losses are non-negative. For power systems studies
under dynamic operating conditions, ramp-rate constraints such as u̇gen,i and u̇load,i
can be included. Further constraints can be imposed on the variables ugen,i , uload,i ,
Ci , xi , ξi , νi and wi to represent the availability and controllability of an arbitrary
generation, storage or load unit. Table 1.2 establishes a set of basic properties iden-
tified by the authors which defines the operational behavior of a power system unit
modeled as power nodes.
Table 1.2: Notation of a Single Power Node [11]
Variable(s) Constraint(s) Implications

ugen,i , uload,i ugen,i = 0 Load
uload,i = 0 Generator
ugen,i .uload,i One-conv.-unit storage
- Two-conv.-unit storage
Ci Ci = 0 Non-buffered unit
Ci >0 Buffered unit
ξi ξi = 0 No external process
ξi ≥ 0 Supply process
ξi ≤ 0 Demand process
ξi , wi ξi = ξdrv,i (t) ∧ wi = 0 Non-controllable
ξi = ξdrv,i (t) Curtailable
ξi arbitrary, wi = 0 Controllable
vi vi = 0 Lossless storage
vi > 0 Lossy storage
u̇gen,i u̇min max
gen,i ≤ u̇gen,i ≤ u̇gen,i Ramp-rate-constr. gen.
u̇load,i u̇min max
load,i ≤ u̇load,i ≤ u̇load,i Ramp-rate-constr. load.
The selection of constraints are clarified in the following:
• The injection or consumption of power from a power node is determined by

the variables ugen,i and uload,i . A pure generation process will perpetually
hold uload,i = 0 and a pure load ugen,i = 0. A storage process which holds
bi-directional conversion system is more particular as either one of the two
variables must always be zero, e.g in an inverter connected battery storage,
or there is no constraints on neither; the conversions can happen in the same
time e.g pumped hydro plant with independent pump and turbine.
• If a power node is capable to store energy, then its capacity is non zero
(Ci > 0). Otherwise, Ci = 0.
• The sign of the external process variable ξi determines its nature; ξi > 0 is
supply e.g sun, wind, fuel; ξi < 0 is demand e.g electrical appliances. If no
external process takes place e.g battery, then ξi = 0
• Constraints on ξi and wi determine the controllability of a unit. If an ex-

ternal time-dependent signal drives ξi (ξi = ξdrv,i (t)), e.g induced by a load
demand, the unit may either be regarded as non-controllable (no load cur-
tailment is possible: wi = 0) or curtailable (no constraint on wi ). ξi is not
externally driven, then the unit is considered controllable and wi = 0 because
the curtailment of a controllable unit is irrelevant.
• If the power node feature storage, then it is either considered lossy (ν > 0)
or not (ν = 0)
• Power production and consumption may be rate-constrained which is re-

flected by bounding their time derivatives
1.6 MPC application in power system dispatch

Weather predictions are not perfect. Real time errors in predicting the future out-
put of variable renewable energy (VRE) are causing troubles for grid operators to
balance the system and will worsen with more VRE in the mix. The traditional
economic dispatch (ED) algorithm is not equipped with a controller that could
predict the changes in the output of wind and solar units because of a change in
weather forecast. Model Predictive Control (MPC) algorithm comes very in handy
as it solves the optimization problem in a receding horizon instead of a single time
step like the ED. The receding horizon defines the time span in which weather and
load demand forecast are available, also known as the prediction horizon length.
This time span will define the measured inputs that are fed into an optimizer with
the goal of minimizing the system cost of operation (economic dispatch). Only
the first optimal decision variables are implemented and are fed in as input to the
second optimization step within the prediction horizon length. The latter is shifted
forward by one step receding into the future. As this shift is made, weather updates
are incorporated, thereby any change in predicted wind speeds or sun irradiation
will be given as input to the now current state, yielding a new dispatch strategy.
The iteration is repeated until the last time step of the simulation length of the dis-
patch operation. To perform an MPC controlled economic dispatch close to reality,
the prediction horizon can be set to 24 hours according to the bids in the day-ahead
market. Weather updates can be performed every second, however this would sig-
nificantly impede the computation time. P. Jonas studied in his master thesis [54]
the effect of the weather update frequency on the dispatch cost. He showed that
the former can be set up to one sixth of the prediction horizon length of 24 hours
without compromising the optimal solution. Based on his results, weather updates
should be done every 4 hours to re-calculate an optimal trajectory in case predicted
wind or solar infeeds have changed. Figure 1.17 illustrates the application of MPC
in power system dispatch.
Figure 1.17: MPC application in power system dispatch.
The example above represents only a sketch of the algorithm solution and no
real solution. One observes the change in the current trajectory of the decision
variables after an update of the future wind forecast. With such an algorithm,
storage plants are optimally cycled as they are discharged prior to an expected
excess intermittent renewable infeed and the use of peaking power plant is reduced
as inexpensive plant can be ramped up prior to supply shortfalls. Chapter 4 will
detail the mathematical formulation of the the economic dispatch and the MPC
model.
1.7 Upgrading the future EU power system

To develop optimally the European power system (EPS), one must assess the value
of the listed flexibility options in figure 1.6, individually but also assess the interde-
pendencies between them. To do so, we analyze the sensitivity of VRE integration
to the linear capacity increase of all possible combinations of flexibility sources.
We choose as base year 2020 because Europe will very likely meet its 20% target
share of RES in final energy consumption [56]. Thus, we assume that the NREAP
of each EU country will be applied by 2020; all additional generation, pump stor-
age,wind and solar units and transmission lines capacity will be installed. This
forms our basic flexibility criteria. We increase the VRE supply fed to the 2020
EPS and analyze the sensitivity of VRE integration to the increase capacity of flex-
ible resources. The optimal upgrade is found just before the integration of high-
RES starts to saturate. The value of flexibility of each source is calculated on the
basis of the additional savings obtained in operating an upgraded EPS under a high
renewable scenario. Due to time constraints and increased complexity of the sys-
tem, we do not evaluate here the additional flexibility electric vehicles and thermal
loads could offer to grid operators. As well, since not all the VRP are equipped
with the innovative power electronic products that ensure their controllability, we
only consider the curtailability option of these units, which can be instantaneously
applied by disconnecting the PV inverters from the grid or pitching the rotor angle
of the turbine blades and shutting down the plant in less than 5 min [57].
Part I
Modeling Framework
30
Chapter 2
Modeling data
We introduce the data we used from various sources (IRENE 40, EurElectric and
IEA [20, 58, 18]), how we structured it and how we scaled it up if needed. Data
consisted of every EU country’s generation mix, pump storage capacity (energy
and power) and the net transfer capacities limits between the countries provided by
ENTSO-E. We explain the rational behind the hourly time series of load, PV, CSP,
wind onshore, wind offshore that were also provided. We clarify how we modeled
hourly hydro (reservoirs and run-of-river) time-series for every EU country based
on the provided data of Switzerland. We present also in this chapter how we derived
the constraints to our optimization problem, mainly the cost constraints and the
ramping constraints.
32
CHAPTER 2. MODELING DATA 33
2.1 IRENE-40
The EU research project IRENE-40 (Infrastructure Roadmap for Energy Networks
in Europe) grouped highly credential industries (ABB, Alstom, ECN, Siemens)
and European research universities (ETH, Imperial, NTUA, RWTH, TUDelft) to
plan the development for the European and Pan-European transmission networks
for the next 40 years [12]. In addition, they evaluate future generation mixes and
load scenarios depending on different paths the EU can take in achieving its 2050
target of reducing CO2 emission by 80 % reduction target. They categorize five
scenarios in the following:
Table 2.1: Overview of the five IRENE-40 demand and generation scenarios [12].
Scenario Short description

1 BAU 80 % CO2 reduction in 2050 not achieved
2 EFFICIENCY Lower electricity demand than other scenarios
3 RES High contribution of RES to 80 % goal
4 DESERTEC Similar as RES but with import from Africa
5 CCS Substantial contribution from CCS to attain 80% goal
To be able to reduce the domestic (Power sector, residential, industry, transport,

agriculture) CO2 emissions of Europe by 80% by 2050, the power sector must be
up to 93 to 99 % decarbonized [23]. Only the Business as Usual (BAU) scenario
in IRENE-40 does not reach such reduction targets. The other four scenarios are
intended to lead to a CO2 reduction of about 90-95% in 2050 compared to 2010
levels [12]. Other than the CO2 emission target, these scenarios are mainly distin-
guished by the electricity demand growth. The latter is depicted in figure 2.1.
Figure 2.1: Final electricity demand for the EU in the five scenarios in TWh per
year [12].
The electricity demand increases from 2010 to 2050 by 10%, 40% and 60%
in the EFFIENCY, BAU and the other scenarios, respectively [12]. Since the RES
scenario is the one with most intermittent share in the mix, 50%, and assumes
the highest rise in electricity demand due to the introduction of electric transport
and heatings [12], we chose to use its databases for the year 2010, 2020 and 2050
to assess the highest flexibilty value for the European power system. Data basis
consisted of hourly load demand, wind (offshore and onshore) and solar (PV and
CSP) infeeds in every EU country. We discuss also how they planned to develop the
transmission network from 2020 onwards and the assumptions they took to change
the generation fleet. Their RES 2050 scenario will be benchmarked in Chapter∼
7 with respect to the VRE integration, the operation cost of the system and CO2
emissions to our proposed power system upgrades.
2.1.1 Load forecasts

To adjust the load curves of every country to economic, political, social and techno-
logical factors, IRENE-40 used historical load data to develop trends and forecast
future deviations in three stages:
• Obtained ENTSO-E annual hourly data from the year 1990 to 2010 for the
27 EU member states, Norway and Switzerland.
• Scale up the country annual load curves for the years 2020, 2030, 2040, 2050
to match the total energy forecast demand of the EU for every decade.
• Adjust the contribution of each month to the annual electricity consumption

in every country by extrapolating the linear trends identified in historical
data. By doing so, they include the ex-ante effects climate conditions, con-
sumer behavior, energy efficiency could have on electricity demand. Figure
2.2 illustrate an example of this calculations made for Italy for the month of
January.
Figure 2.2: Monthly consumption (January) as a percentage of the annual con-

sumption for Italy - historical data, extrapolated data, trends [12].
2.1.2 Wind energy

IRENE-40 obtained wind hourly power generation per country from the TradeWind
project, which used wind speed data provided from the Reanalysis global weather
model [12]. This data covered all Europe and was only available at six-hourly in-
tervals. TradeWind interpolated to hourly intervals and adjusted the wind speed,
the hub height and the wind shear exponent to the terrain in which wind farms were
located (Lowland, Upland, Offshore) [13]. By doing so, they created two different
wind profile time series for wind onshore and wind offshore in every EU country.
After comparing the years 2000-20006, they identified year 2004 to be the most
challenging year for wind integration in Europe because of the high winds, espe-
cially in Germany, the country with the highest amount of installed wind power
capacity [13]. From this wind profile, IRENE-40 calculated power ratios and mul-
tiplied it with the installed capacity for a given year and scenario [12]. Therefore,
the 2010 wind energy production in IRENE-40 scenario does not necessarily match
the 2010 actual production. For example, according to IRENE-40, Germany has
produced in 2010 56 TWh of wind energy whereas, according to the ENTSO-E
measurements, it has produced 36 TWh on that year with the same capacity [59].
However, what is relevant for our research is the variability of the time series and
the differentiation between wind onshore and wind offshore power series. Figure
2.3 exemplify the validity of the technique used by TradeWind to simulate wind
power production in Denmark in year 2000 since the Reanalysis curve follows the
trend of the actual measurement. Figure 2.4 portrays the difference between the
wind onshore and wind offshore power series in Germany and the increase in vari-
ability with an increase in capacity.
Figure 2.3: Measured historical wind power production in western Denmark for
third week in 2000 compared to simulated data, using Re-analysis wind data [13].
Figure 2.4: Simulated wind onshore and offshore power infeed on the third week
of January in Germany for different reference years [12].
2.1.3 Solar energy

Solar hourly data were calculated on the basis of the satellite data available at
the SODA database. The photovoltaic (PV) infeed follows the normal distribution
curve with mean at 0 on a sunny day. IRENE-40 assumes that concentrated solar
power (CSP) plants are equipped with an 18 hour heat storage systems so that the
daily Solar power output from CSP plants is flattened. They transform in this way
the hourly solar variation into daily variations. This assumption is technically valid
as in May 2011, a CSP plant was commissioned in Spain to offers 24 hours of solar
electricity by utilizing molten salt storage [60]. Figure 2.5 display the simulated
solar infeed in Spain on a sunny day during the first week of January for different
decades.
Figure 2.5: Simulated PV and CSP infeed on the first week of January in Spain for
different reference years [12].
2.1.4 Net transfer capacities

IRENE-40 provided us with NTC values in Europe from the year 2010 till 2050.
For 2010, they averaged the summer and winter NTC values of this year published
by ENTSO-E. These were derived as explained previously in section 1.2.4. The
2010 values are shown in the appendix (D.2). For 2020, they added the network
reinforcements planned by the ENTSO-E in [61] to the NTC values in 2010. They
distinguished between AC and DC network reinforcements. In case of DC, they
assumed that the reported physical capacity is equal to a pure NTC reinforcement.
In case of AC, they multiplied the proposed physical capacity increase by the ra-
tio of the existing NTC and physical capacity for each interconnection in 2010.
For 2050, they assume a minimum target for transmission capacity as fraction of
installed generation capacity for each EU member state. Thereby, 20% of NTC ca-
pacity as fraction of total installed generation capacity and 10% of NTC capacity
as fraction of intermittent (i.e. wind + PV) generation capacity is required to be
installed per country. To get country to country NTC capacity estimation in 2050,
they multiply the total NTC need of a country in 2050 by fractions of its inter-
connections with its neighbors back in 2020. In our study, we do not follow their
approach in estimating the NTC needs in the future. We fixed our base simulation
to the power system in 2020 and assesses the needs for flexibility based on a opti-
mal power system dispatch. We benchmark in Chapter 7 our NTC development to
the IRENE-40 one.
2.1.5 EU generation mix

We received data from IRENE-40 project covering the generation mix of every EU
country for the year 2010 till 2050. A sample of selected countries is shown in
figure 2.6.
Technology*(MW) Austria Belgium Bulgaria Czech*Republic

Renewable*energy*(no*pumping*included) 14818 3033.3 3449 4206
Hydro&(pump&storage&excluded) 8235 112.3 2090 1047
Hydro&(pumping&systems) 4285 1065 1014 1150
Wind&onBshore 1011 703 336 243
Wind&offBshore 0 185 0 0
Solar&PV 90 350 9 1650
Solar&CSP 0 0 0 0
Marine&(Tidal,&waves) 0 0 0 0
Geothermal(Slow&Gen) 1 0 0 0
Biomass&and&Waste&fired&(Slow&Gen) 1196 618 0 116
Biomass&CoBfiring 1099 498 0 3
Biomass&StandBalone 97 106 0 113
Waste&StandBalone 0 14 0 0
Thermal*power 6833 9215 6233 10018
Solids*fired(Slow*Gen) 1913 2004 5365 8798
Combustion&technologies 1913 2004 5365 8798
Hard&coal 1583 2004 1869 1598
Lignite 330 0 3496 7200
IGCC 0 0 0 0
Gas*fired 3887 6271 868 1136
CCGT&(natural&gas) 1283 3534 0 492
CHP&Gas 680 531 313 16
GT&and&Gas&boilers&(natural&gas) 1864 1877 525 579
IC&engines&(Derived&gases) 60 329 30 49
Oil*fired 1033 940 0 84
Nuclear*energy 0 6040 2000 3902
Figure 2.6: IRENE-40 generation mix in 2010 of Austria, Belgium, Bulgaria and
Czech Republic [12].
For both years 2010 and 2020, since the EU is supposed to meet the "EU 20-
20-20 goals", the figures listed in the generation mix are in line with the NREAP
figures [12]. For the RES scenario, IRENE-40 assumes that renewable technolo-
gies will replace conventional power plants (coal, oil) even before technical end-
of-life. Large clustered offshore and onshore wind farms will be deployed in the
northwest of Europe, solar and wind in the south, and hydropower and biomass in
central and norther Europe. Renewable energy sources amounts to 80% of total
electricity generation. A constant growth rate per renewable electricity source is
assumed for the period 2020-2050. Nuclear capacity decrease from 140 GW to
155 GW and the use of solid and oil fired fuels decreases from 45% in 2020 to
8% of electricity generation in 2050. Figure 2.7 depicts the evolution of the EU
generation mix forecast by IRENE-40 for the RES scenario.
Figure 2.7: IRENE-40 RES scenario: EU generation mix evolution from 2010 to
2050 [12].
In our study we use the country load forecast data and apply the same de-
ployment of renewables in 2050 as in the RES scenario of IRENE-40. We do not
however apply the change of the conventional fleet as we keep it fixed to 2020
to study the flexibility need in ramping requirements in a high-share scenario and
study the sensitivity of CO2 emissions to a change in the 2020 EU generation mix.
2.2 EURELECTRIC
2.2.1 Detailed hydro capacity data
EURELECTRIC publishes in [58] detailed data on the hydro capacities installed in
every country; Hydro reservoirs, run of river and pump storage plants. We include
therefore their data to the IRENE-40 one. Vital to our model and to the integration
of renewable energy1 was the aggregated pump storage energy reservoir capacities
per country obtained form the report [8] shown in figure 1.11. Also, based on
the information they published regarding start-up time and ramp rates of single
conventional generators that was shown in figure 1.9, we derived equivalent ramp
rates to a fleet of generators for different operation scenarios.
1
Chapter 5 will elaborate on this issue
2.2.2 Equivalent ramp rates of a pool of conventional generators
Table 2.2: Start up time and ramping constraints of conventional power plants [7].
Power plant Nuclear Coal Gas

Start up time "Cold" 40h 8h 2h
Ramp rate 5 %/min 2%/min 4%/min
75 %/15min 30%/15min 60%/15min
Time to reach full capacity 20 min 50 min 25 min
Minimal possible load 50 % 40 % 40 %
This table exemplifies the fact that the conventional fleet on an hourly basis is very
flexible if it is already turned on. We make two key assumptions in deriving an
approximative hourly ramp rate constraint for a pool of generators:
• Any ramping down operation will happen at a load operation > 50%. Thereby
no ramping down constraint is applied and any plant could shutdown in less
than an hour.
• Any ramping up operation is happening at a load operation < 50%. Start up

time of the plant need to be taken into account.
We identify four scenarios for the state of the fleet of a country. By fleet we mean
the aggregation of all nuclear power plants, coal and gas respectively. By state we
mean if the plant is turned on or off.
1. All the fleet is turned off and need to start up from cold
2. 75 % of the fleet is turned off and need to start up from cold

The following table displays equivalent ramp rates calculated on the basis of the
stat up time, the state of the fleet and the ramping capability of the plant in question
Table 2.3: Equivalent ramp rates of a pool of aggregated conventional generators.
Scenario Nuclear (%/h) Coal (%/h) Gas (%/h)

1 2.48 11.32 41.38
2 3.3 14.63 52.17
3 4.9 20.7 70.6
4 9.7 35.3 109.1
The nuclear power plant calculations for scenarios 1 and 2 are shown below:
100%
(1) = 2.48 %/h (2.1)
40h + 20min
60min
100%
(2) = 4.9 %/h
40h ∗ (1 − 0.25) + 60min
20min
2.3 Swiss Federal Office of Energy

2.3.1 Run-of-river and hydro reservoir hourly power infeed
The Swiss Federal Office of Energy (SFOE) published every year detailed hydro
data [62].Figure 2.8 illustrates the time-series that the PSL laboratory have ag-
gregated for all the run-of-river plants and hydro reservoirs plants in Switzerland.
Table 2.4 groups the aggregated power production data for each type of plant in
Switzerland.
Figure 2.8: Run-of-river and hydro reservoir hourly infeed in 2010 in Switzerland.
Table 2.4: Switzerland hydro mix yearly production [17].

Type Capacity (MW) Energy (GWh) Capacity factor
Hydro reservoir 8073 16840 0.238
Run-of-river 4224 17556 0.474
2.3.2 Simulating hydro infeed in rest of EU

Based on Switzerland’s data we simulated the run of river and hydro reservoir
infeed in the rest of the EU countries by following these steps:
1. Assume similar capacity factors for each type of hydro plant in all EU
2. Compare the total energy produced2 in each country per each type of tech-
nology to values published in IEA electricity report 2010 [14]
3. If the values match
• RoRhourly,Spain = RoRhourly,Swiss ∗ (RoRtotal,Spain /RoRtotal,Swiss )

• Similar calculation for reservoir infeed.
4. If the values do not match
• Modifiy capacity factor iteratively until you reach a small difference in

values compared.
2.4 IEA
2.4.1 Marginal cost of production of power plants
The joint report by the IEA and the OECD Nuclear Energy Agenecy presents the
latest data concerning electricity generation costs in 2010 [18]. The data spans a
wide variety of fuels and technologies, including coal and gas, nuclear, hydro, on-
shore and offshore wind, biomass, solar and combined heat and power plants [18].
The electricity generation costs were obtained from surveying 200 plants in 21
countries, several industrial companies and organizations. We chose as reference
the industrial values since we assumed the EU electricity market to be unified. If
there was no assigned industry value, we used a country value and assumed it ap-
plies for the rest of thee EU countries. We present here an overview of the marginal
cost we have assigned to each technology. The marginal cost include the fuel costs,
the operation and maintenance cost and the CO2 emission cost. The latter is evalu-
ated at a CO2 market price of 10e/MWh and not 30$/MWh like in [18]. We found
that this value does not reflect the CO2 market price of today. Plant specific CO2
emissions per MWh of electricity produced are obtained from [19]. The financial
conversion rate assumed is 1e= 1.3$.
2
multiply capacity installed by capacity factor of technology
Table 2.5: Marginal Cost of Production of conventional power plants [18, 19].
Power Plant Fuel import Emissions Efficiency Fuel cost O&M Carbon cost Total cost
(e/GJ) (tCO2 /MWh) (%) (e/MWh) (e/MWh) (e/MWh) (e/MWh)
Nuclear 0.72 - 36 7.2 9.1 - 16.3
Oil 13.4 0.62 45 107.2 4.3 6.2 117.7
Gas turbine 7.5 0.37 45 60 4.14 3.7 67.8
CCGT 7.5 0.37 58 46.5 3.1 3.7 53.3
Black Coal 2.8 0.89 45 22.3 3.9 8.9 35.1
Lignite 1.3 0.94 43 10.5 4.04 9.4 23.9
Oil prices were not found in [18]. We averaged the import price of oil in Spain,
Italy and France we found in [63]. The next table presents the marginal cost of
production from renewable energies.
Table 2.6: Marginal Cost of Production of renewable power plants [18].
Power plant Fuel cost O&M Carbon cost Heat credit Total cost
(e/MWh) (e/MWh) (e/MWh) (e/MWh) (e/MWh)
Wind Onshore - 26.8 - - 26.8
Wind Offshore - 37.4 - - 37.4
Solar PV - 22.5 - - 22.5
Solar CSP - 28.2 - - 28.2
Biomass-CHP 12.3 9.3 2.4 17.1 6.9
Biomass Solid3 53.1 3.5 - - 56.6
Municipal Waste4 - 37.9 22.1 33.9 26.1
Geothermal - - 4.21 - 4.2
Hydro - 3.9 - - 3.9
Pump Storage - 8.12 - - 8.1
The operation and maintenance cost of PV panels is debatable since PV plants

are stationary and do not require more than dust removal or cleaning from eventual
snow or dirt. In our simulation, we chose to dispatch the power system according to
a scenario where the O&M cost of wind and solar plants is subsidized and therefore
equal to zero, and another scenario, where renewables must compete in the market
with conventional power plants according to the derived cost data. The difference
in results for each scenario is further clarified in Chapter 6.
2.4.2 Cost projections for 2020

To project the cost of electricity generation in 2020, we assume:
3
Netherlands value
4
Czeque Republic value
1. Gas, coal, oil prices increases in 10 years by 60%, 70% and 50% while
uranium price does not change according to the IRENE-40 forecast [12].
2. Conventional power plant efficiency improves by 2%.
3. O&M cost of nuclear, wind onshore, wind offshore, solar PV, solar CSP
decreases by 7%, 5%, 8%, 22%, 21% according to the VGB report [64].
4. CO2 emission price is 15 e/ton instead of 10 e/ton.
Table 2.7 summarizes the forecast of the marginal cost of power production in
2020.
Table 2.7: Forecast of the marginal cost of power plants in 2020.
Marginal cost of production 2010 2020

(e/MWh)
Nuclear 16.3 15.6
Oil 117.7 149.3
Gas turbine 67.8 98.5
CCGT 53.3 74.1
Black coal 35.1 49.3
Lignite 23.9 38.3
Wind Onshore 26.8 25.5
Wind Offshore 37.4 34.5
Solar PV 22.5 17.6
Solar CSP 28.2 22.3
Biomass CHP 6.9 20.8
Biomass solid 56.6 70.6
Municipal waste 26.1 26.1
Geothermal 4.2 4.2
Hydro 3.9 3.9
Pump storage 8.1 8.1
2.5 Final structure of one country bus

To reduce the computational time, the weighted average cost per country of black
coal and lignite technologies is computed to represent the cost of coal power plants.
Mathematically, this is equivalent to:
HCCap,i (MW) ∗ HC p (e/MWh) + LigCap,i (MW) ∗ Lig p (e/MWh)

Coali (e/MWh) =
(HCCap,i (MW) + LigCap,i (MW))
Similarly, the weighted averaged cost per country of biomass-CHP, biomass

solid and Municipal Waste is computed to represent the cost of Biomass technolo-
gies. Figure 2.9 displays how we structured the fleet of one country according to
the controllability of the plant, its rampability, its efficiency and its marginal cost
of production. The ramping constraints used in this structure are the ones derived
when we assumed the TSO will keep 50% of the fleet turned on.
Country Fleet !
Constrained Controllable
Curtailable Generation! Generation (up ramping)!
Instantaneous! Slow Ramp! Medium Ramp! Fast Ramp! Instantaneous!

100%\hr! 5%/hr! 21%/hr! 71%/hr! 100%/hr!
PV ! Nuclear! Geothermal! Gas! Hydro reservoir!

ξ infeed! η = 38%! η = 20%! η = 47%! ξ infeed!
Wind Onshore! Solar CSP! CCGT! Run of river!

ξ Infeed! ξ infeed! η = 60%! ξ infeed!
Wind Offshore! Biomass! Oil! Pump Storage!

ξ infeed! η = 47%! η = 47%! η = 85%!
Coal!
η = 47%!
Figure 2.9: Country fleet structure according to controllability, ramping capability,

efficiency and cost of power plants.
We identified in this structure the operational flexibility of the generation fleet

of every EU country. (ξ) infeed corresponds to the power node notation of an
external supply. Hydro and Solar CSP technologies have an external supply but are
controllable generation technologies under certain ramp rates as we have explained
in sections 1.2.2 and in 2.1.3. PV, Wind onshore and offshore are not controllable
but curtailable. The rest of the fleet is conventional generators that are controllable
under certain ramp rates and efficiency. The next chapter explains how we modeled
the EU power system based on this structure and using the Power Node method.
Chapter 3
Modeling power nodes
We clarify how we modeled each type of generators, storage and loads according
to their controllability and operational constraints using the Power Node method.
We group them into cluster of nodes feeding and consuming electricity from one
bus, which represents one EU country. The buses are interconnected via power
lines and exchanges are constrained to the net transfer capacities. A system view
of the interconnected EU power system modeled as Power Nodes is illustrated.
48
CHAPTER 3. MODELING POWER NODES 49
In our study, we model the 27 EU member states, Switzerland and Norway,

thereby we have 29 buses. Each bus consists of a total of 15 nodes: one load which
draws power from the bus, 13 generator nodes which inject power, 1 storage node
which is able to do both operation. We categorize them according to their type
T, which is determined by their power node equation, constraints, parameters and
decision variables:
1. T1 : Conventional Load
2. T2 : Controllable Plants
3. T3 : Non-buffered curtailable renewable plants
4. T4 : Hydro reservoir
5. T5 : Pump storage plant
3.1 Conventional load

The aggregated electricity demand of every EU country is lumped into one Power
Node per country. The node is considered to have no storage capacity (C=0). Load
demand is driven by an external profile (ξ = ξdrv,i (t) ≤ 0) and can be shedded as
last control eligible action to always allow the solver to find a feasible solution.
We assume a load draws power from the grid without losses, therefore ηload = 1.
According to the general power node equation introduced in 1.5, the Power Node
equation that describes a conventional load of one nation N becomes:
N : ξ1 − w1 = −uload,1 (3.1)
Figure 3.1 illustrate the external load demand ξdrv (t) of France in 2010 which
we obtained from the IRENE-40 database.
Figure 3.1: External demand process ξdrv (t) representing the aggregated electricity
demand in France in 2010.
The maximum load in France during that year amounted to 89.6 GW. This
value constrains the maximum power this node can absorb from the bus in one
hour. It defines as well the maximum curtailment possible in the worst case sce-
nario. Table 3.1 summarize the characteristics (equation, constraints, parameters,
decision variables) of the conventional load Power Node i=1 in France, which is
applied similarly for the rest of the EU countries.
Table 3.1: Characteristics of the Power node i = 1, which represents the aggregated
load demand in France.
Power Node equation

ξ1 − w1 = −uload,1
Constraints on a non-buffered load with curtailable demand
ξ1 = ξdrv,1 (t) ≤ 0 C1 = 0
ugen,1 = 0 w1 = 0
Power Node parameters
umax
load,1 = 89.6 GW load,1 = 0
umin
u̇load,1 = 89.6GW
max ẇmax
1 = 89.6 GW
Decision variables1
uload,1 w1
3.2 Generation with curtailable supply

3.2.1 Non-buffered renewable plants
Non-buffer (C = 0) intermittent generation units do not consume any power from
the grid (uload = 0) and are curtailable (wi ≥ 0). Their maximum power output
can be almost instantly curtailed. No ramping down constraint on is implemented.
They directly feed in electric output from the power node supply domain to the
grid domain therefore ηgen =1. The power node numbering notation of non-buffer
renewable plants follows the list below
Table 3.2: Non-buffered renewable Power Node numbering list.
Power Node Plant type

2 Solar PV
3 Wind Onshore
4 Wind Offshore
5 Solar CSP
6 Run of river
Figures 2.5, 2.4, 2.8 showed external supply profile ξdrv,i (t) ≥ 0 for these tech-
nologies. Table 3.3 summarize the characteristics of the PV power node, which is
similarly applied for the rest of the non-buffer renewable power nodes.
1
Determined by the optimizer. Chapter 4 will elaborate on this issue.
solar PV installations in any EU country.
Power Node equation

ξ2 − w2 = ugen,2
Constraints on a non-buffered generator with curtailable supply
ξ2 = ξdrv,2 (t) ≥ 0 C2 = 0
uload,2 = 0 w2 ≥ 0
gen,2 = max(ξdrv,i (t))
umax gen,2 = 0
umin
u̇gen,2 = max(ξdrv,i (t))
max ẇmax
2 = max(ξdrv,i (t))
Decision variables
ugen,2 w2
3.2.2 Buffered hydro reservoir

Hydro reservoirs are characterized by their large capacity (C) to store energy by
impounding water behind a dam at a high altitude. Water leakage or evaporation
efffects in the storage reservoir are neglected (v = 0). Hydro reservoirs gener-
ate electricity in the same way pump storage plants do2 . The externally driven
power series ξdrv (t) ≥ 0 of hydro reservoirs in Switzerland was illustrated in fig-
ure 2.8. Table 3.4 summarize the hydro reservoir power node characteristics in
Switzerland. They applies similarly to other EU countries with hydro reservoirs
capacities. To find the hydro reservoir energy capacity in every EU country, we
scaled Switzerland’s official value [17] according the the ratio of hydro reservoirs
power capacities between the country in question and Switzerland.
2
refer to 1.2.2
hydro reservoirs installations in Switzerland.
Power Node equation

gen,7 ugen,7 + ξ7 − w7
C7 ẋ7 = −η−1
Constraints on a buffered generator with curtailable supply
ξ7 = ξdrv,7 (t) ≥ 0 v=0
uload,7 = 0 w7 ≥ 0
umax
gen,7 = 2.5 GW gen,7 = 0
umin
u̇gen,7 = 2.5 GW
max ẇmax
7 = 2.5 GW
ηgen,7 = 1 C7 = 8780 GWh3
Decision Variables
ugen,7 w7
3.3 Generation with controllable supply

3.3.1 Non-buffered conventional plants
Non buffer (C = 0) generation units do not consume any power from the grid
(uload = 0) and are controllable (wi = 0). They can be ramped up according to
the constraints we derived in table 2.3. The power node numbering notation of
non-buffer renewable plants follows the list below
Table 3.5: Non-buffer conventional Power Node numbering list.
Power Node Plant type

8 Gas Turbine (Single)
9 CCGT
10 Oil
11 Coal
12 Biomass
13 Geothermal
14 Nuclear
Table 3.6 groups the characteristics of a single gas turbine power plant power
node, which is applied similarly to the rest non-buffer conventional power plants.
3
Source [17]
single gas turbine power plants installed in any EU country.
Power Node equation

ξ8 = η−1
gen,8 ugen,8
Constraints on a non-buffered controllable generator
ξ8 ≥ 0 C8 = 0
uload,8 = 0 w8 = 0
umax
gen,8 = Capacity of country gen,8 = 0
umin
up
u̇gen,8 = 70 %/hr 4
gen,8 = Capacity of country
u̇down
ηgen,8 = 0.47
Decision Variables
ugen,8 ξ8
3.4 Storage technologies

3.4.1 Pump storage hydro plant
Unlike the large hydro reservoir plants, the dispatch of pumped storage hydro
plants do not depend on a water influx (ξi ≈ 0). It is purely controllable and
depend on the need to store or produce electricity by pumping or turbining water
between a lower and upper basins. The energy capacity C of every EU country was
determined by EURELECTRIC in [8] and shown in figure 1.11.
Table 3.7 summarizes the characteristics of the pump storage power node in
Switzerland, which is applied similarly to the rest of the EU countries.
4
Value derived in table 2.3 for scenario 3
Table 3.7: Characteristics of the Power node i = 15, which represents the aggre-
gated pumped storage plants installed in Switzerland.
Power Node equation

C15 ẋ15 = η−1
load,15 uload,15 − ηgen,15 ugen,15
−1
Constraints on a pump storage plant

C15 = 369 GWh 5 ugen,15 .uload,15 = 0
ξ15 = 0 w15 = 0
v15 = 0
umax
gen,15 = 1.38 GW gen,15 = 0
umin
u̇gen,15 = 1.38 GW
max ηgen,15 = ηload,15 = 0.85
Decision Variables
ugen,15 uload,15
3.5 An interconnected European power node system

We present in figure 3.2 a system view of the western European power system
modeled as interconnected buses, each with 5 types of power nodes feeding to or
absorbing power from the bus. The buses can exchange power limited to the net
transfer capacities assigned to each power line.
6
Source [8]
Own extraction = Extraction of energy within the
entity
Stock change = Addition/Extraction of energy from
stocks
44

System Representation
DK
GB
…..…..
T2 T3 T4 T5 T1
…..…..
T2 T3 T4 T5 T1
PO
T2 T3 T4 T5 T1
DE
<>
FR
<>
T2 T3 T4 T5 T1 …………..
<>
<>
T2 T3 T4 T5 T1 CZ
T2 T3 T4 T5 T1
<>
<>
<>
AU
<>
CH
……..
T2 T3 T4 T5 T1 T2 T3 T4 T5 T1
……..
<> <>
…..
…..
T1..T5 Power Node Types Energy flow to/from Bus Curtailment Energy supply/demand <> NTC Import/Export Limit
Friday, May 10, 13 Farid Comaty 11
Figure 3.2: System view of interconnected power nodes representing the western
European power system.
With the power node modeling method, all the different flexibility measures
have been unified in one system and can be assessed. The goal is to now dispatch
this system with the least possible cost in a unified European electricity market and
analyze how does the existing flexibility of the European power systems reacts to
an increasing share of intermittent wind and solar energy.
Chapter 4
The dispatch simulator
We discuss the structure and functionality of the MATLAB-based optimizer which

we used to minimize the European power system operation cost. The backbone
of the simulator is the YALMIP toolbox which is used to translate the power node
model into a SDP1 optimization problem and calls the IBM CPLEX solver to find
the optimal economic dispatch. The mathematical formulation of the optimization
problem is clarified and we explain the rational behind the MPC controller param-
eters that have been heuristically chosen to optimally cycle storage reservoirs. We
implement a sensitivity analysis to find the value of each flexibility parameter in a
European power system under high-shares of renewable energies.
1
Semidefinite Programming
58
CHAPTER 4. THE DISPATCH SIMULATOR 59
4.1 Structure of the simulation tool

In this section, the MATLAB software tool developed by the Power Systems Lab-
oratory (PSL) of ETH is outlined. To decrease computational time, the structure
of the software is coded such that parallel computing is allowed. To stay concise,
we do not illustrate all functions. The full code can be found on the attached CD-
ROM. Figure 4.1 depicts the high-level structure of this tool and the function of
each method is pinpointed.
Figure 4.1: Basic structure of the MATLAB-based simulation tool.
A maximum of four MATLAB interfaces (4 simulations) can be ran in paral-

lel due to computer memory RAM constraints. The power node matrix in De-
fine_powernode_topology was fine tuned to regroup all the operational data of
each bus as structured in Chapter 2. The Power nodes are constructed in Con-
struct_Powernode as described in Chapter 3. The continuous differential power
node equations system is discretized in Build_Dispatch. We explain the latter next
as well as the mathematical formulation objective function and constraints coded
in Solve_Dispatch and the heuristic MPC controller parameters we defined in the
power node matrix to allow for a dispatch close to reality.
4.2 Discretization of the power node system

The single continuous differential power node equations described in Chapter 3 can
be regrouped in a system of equation. Below is an example of the set of equations
that represent one bus. The nomenclature used corresponds to the type of power
node identified in Chapter 3:
ξ1 − w1 = −uload,1 (4.1)
ξ2 − w2 = ugen,2
ξ3 − w3 = ugen,3
ξ4 − w4 = ugen,4
ξ5 − w5 = ugen,5
ξ6 − w6 = ugen,6
C7 ẋ7 = −η−1
gen,7 ugen,7 + ξ7 − w7
ξ8 = η−1
gen,8 ugen,8
ξ9 = η−1
gen,9 ugen,9
ξ10 = η−1
gen,10 ugen,10
ξ11 = η−1
gen,11 ugen,11
ξ12 = η−1
gen,12 ugen,12
ξ13 = η−1
gen,13 ugen,13
ξ14 = η−1
gen,14 ugen,14
C15 ẋ15 = η−1
load,15 uload,15 − ηgen,15 ugen,15
−1
A continuous-time space model with input vector u and state vector x can be
formulated. The vector u groups the decision variables the solver needs to optimize
and the externally driven time series supplied
u =[uload,1 , w1 , ugen,2 , w2 , ugen,3 , w3 , ugen,4 , w4 , ugen,5 , w5 , ugen,6 , w6 , ... (4.2)

ugen,7 , w7 , ugen,8 , w8 , ugen,9 , w9 , ugen,10 , w10 , ugen,11 , w11 , ugen,12 , w12 , ...
ugen,13 , w13 , ugen,14 , w14 , ugen,15 , uload,15 , ξ1 , ξ2 , ξ3 , ξ4 , ξ5 , ξ6 , ξ7 , ]
x groups the dynamic state variable of the storage nodes with capacity C
x = [x1 , x7 , x15 ] (4.3)
The state space model is further discretized in sampling time t and represents
the dynamics of an electric power system dispatch modeled as power nodes
x(t + 1) = A.x(t) + B.u(k) (4.4)

A and B matrices are computed in the method Build_Dispatch.
4.3 Economic dispatch optimization

To perform an hourly dispatch of the European interconnected power node system2
in the most efficient and profitable way, an objective function need to be minimized.
The latter is constrained to physical and operational limits that derives from gener-
ating electricity, storing it or transmitting it. With the implementation of an MPC
algorithm that allows to optimally cycle storage plants over a prediction horizon N,
the objective function J, at time step t, can be formulated as follows:
l=t+N−1
X
minu J(t) = ((x(l) − xre f )T · Q x · (x(l) − xre f )) (4.5)
l=t
+ u(l)T · Qu · u(l) + Ru · u(l)
+ δu(l) · δQu · δu(l)
s.t. (a) x(l + 1) = A · x(l) + B · u(l)

(b) 0 ≤ xmin ≤ x(l) ≤ xmax ≤ 1
(c) 0 ≤ umin ≤ u(l) ≤ umax ≤ 1
(d) 0 ≤ δumin ≤ δu(l) ≤ δumax ≤ 1

(e) ξ(l) = ξdrv (l, T )
X i=1,2,3,4,5,6,7X i=1,2,3,4,5,6,7
X X
(f) ugen,B (l) − uload,B (l) = Pex,B − Pim,B
(g) 0 ≤ Pex,B ≤ NTCex,B
(h) 0 ≤ Pim,B ≤ NTCim,B
(a − h) ∀l = { t, ..., t + N − 1}
The system is composed of 29 interconnected buses, with 15 power nodes

per bus and two decision variables per power node. Therefore, each optimiza-
tion step consist of finding the least cost dispatch of 870 (29x30) variables such
that the above constraints are satisfied. Constraint (a) expresses the discretized set
of the system of linear Power node equations described previously, (b) the state
of charge of a hydro reservoir or pump storage plant is normalized and bounded,
(c) the input-variables are non-negative and bounded, (d) the input-variables are
2
refer to figure 3.2
rate-constrained, (e) the demand-supply process of the power nodes i ∈ B =

{1, 2, 3, 4, 5, 6, 7} are externally driven, (f) the power balance of each area is equal
to the difference between all power export from this area to its neighborings and
its power imports from the latter, (g) the aggregated power exports per bus is con-
strained to the sum of respective NTC export limits, (h) the aggregated power im-
ports per bus is constrained to the sum of respective NTC import limits.
As was explained in section 1.6, the first optimal set of decision variables is
computed over a prediction of 24 hours and applied for hour 1 and is then fed back
as additional input to the next optimization at hour 2, which is in turn computed
over a horizon of 24+1 hours. Hence, at each hourly step t, the day-ahead cost
terms are calculated and summed up iteratively to obtain the total dispatch cost
of the European power system. We discuss next the optimization strategy and the
penalty factors Q x , Qu , Ru and δQu we assigned to steer the dispatch behavior of
the MPC controller close to reality.
4.3.1 Characteristics of the optimization strategy

The optimization strategy defines how we set the MPC controllers (penalty factors)
to influence the dispatch behavior of the CPLEX solver. In chapter 2, we derived
the marginal cost of electricity generation and storage, which only controls the
decision of generating or storing electricity. To control the decision of curtailing
renewable output, shedding load, or ramping conventional generators, we heuristi-
cally chose the cost associated to these control actions. To understand our rational,
we clarify first the dispatch strategy inspired from [54]:
1. Load shedding is used as last resort to balance supply and demand within
one area; that is when wind, solar and hydro energy infeeds are insufficient,
pump storages are empty, conventional plants are running at full capacity
under ramping constraints, power imports line are fully loaded.
2. In case priority infeed is given, curtailment of intermittent generation is for-

bidden as long as storage energy reservoir capacities are available to absorb
the excess energy.
3. In a economic optimum dispatch, curtailment of intermittent generation is

allowed if the reduction in balancing cost outweighs the benefit of prioritiz-
ing renewable energy i.e low-cost inflexible units are ramped up prior to a
predicted shortfall at the expense of curtailing the present renewable energy
infeed instead of integrating the current infeed and dispatching high-cost
flexible plants at the coming shortfall.
4. The pump storage plants can either store or supply electricity at once, thereby,
in the case the reservoirs are full and there is a high renewable infeed pre-
dicted, the energy will be curtailed and not cycled through a lossy storage.
Such a strategy reduces the wear and tear of storage plant equipment.
5. Generation and storage unit dispatch should avoid sequences of excessive up

and down ramping to spare mechanical parts from unnecessary abrasion and
fatigue.
4.3.2 MPC controller parameters

4.3.2.1 Storage controller
The first term in equation 4.5 expresses the quadratic cost of any storage deviation
from a reference set point. In order to always have filled storage that can back
up any lower than predicted renewable infeed, the reference state of pump storage
plants is set to xref,i=15 = 1. Only power node 15 (pump storage plant) has a set
reference storage state, power node 7 (Hydro reservoir) should not be penalized
from any deviation and should be dispatched solely according its marginal cost of
production3 . The following penalty matrix Q x per bus result:
Qx = diag(zeros(0, 29), 10) (4.6)

where 10 is chosen heuristically by PSL to represent a quadratic storage cost in
Eur/MWh2 . With experience, PSL have found that such value allows for storage to
be dispatched uniformly prior to an excess of renewable energy and not abruptly.
4.3.2.2 Generation and Curtailment controller

The second term in equation 4.5 represent the quadratic generation cost. The en-
tries in the penalty matrix Qu should be the quadratic fuel coefficients costs that
are given by design manufacture when determining the heat rate of a conventional
plant [65]. However, to reduce the complexity of the optimization problem, we
consider the cost function of generators to be linearly dependent on the sum of fuel
cost, operation and maintenance cost and CO2 emission cost. Therefore we set the
penalty matrix Qu per bus as:
Qu = diag(zeros(1, 30)) (4.7)

The configuration of the third term of equation 4.5, the Ru matrix, depends on
the policy we choose to apply for the simulation:
• Priority infeed with a subsidy for the O&M of VRPs.
• Economic dispatch with a subsidy for the O&M of VRPs.
• Pure economic dispatch where VRPs compete with conventional power plants.
3
One could set the reference state to a reserve value that is used for agriculture purposes and not
solely electricity generation
Priority infeed with a subsidy for the O&M of VRE plants
In this policy scenario, the curtailment cost of VRE is set extremely high (1000
€/MWh) to force the optimizer to integrate VRE as no reduction in balancing cost
could outweigh the benefit of integrating intermittent generation since curtailing it
is very costly. Load curtailment cost is set 10 times higher than VRE curtailment
to force the optimizer to use this flexibility option as last resort. The O&M of
variable renewable output plants identified in table 2.7 is subsidized so it drops to
0 and VRE is considered as a free source of energy. The 2020 marginal cost of
each power node derived in table 2.7 is further inputted.The penalty matrix Ru per
bus becomes
Ru = diag(0, −10000, 0, 1000, 0, 1000, 0, 1000, 0, 1000, (4.8)

3.6, 0, 3.6, 0, 98.5, 0, 74.1, 0, 149.3, 0, 47.6, 0,
27.4, 0, 4.2, 0, 15.6, 0, 8.1, 8.1).
Economic dispatch with a subsidy for the O&M of VRE plants
Under this policy, we remove the priority infeed constraint and allow for a true cost
minimization as highlighted in point 3 of the optimization strategy. The penalty
matrix Ru per bus becomes
Ru = diag(0, −10000, 0, 0, 0, 0, 0, 0, 0, 0, (4.9)

3.6, 0, 3.6, 0, 98.5, 0, 74.1, 0, 149.3, 0, 47.6, 0,
27.4, 0, 4.2, 0, 15.6, 0, 8.1, 8.1).
Since the marginal cost of renewable plants and its curtailment are set free, the
optimizer will never cycle a surplus of energy through storage conversion losses
because the latter comes at a cost of 16.2 €/MWh for both operation.
Pure economic dispatch where VRPs compete with conventional power plants
In case the O&M cost of wind and solar power plant is no longer subsidized, the
curtailment cost of these unit can no be set to zero. If so, then any excess renewable
energy will be curtailed and not stored even though the storage reservoir capacities
might be available. Therefore the first lower bound to renewable curtailment cost
should be the marginal cost of storing 1MWh of electricity using a pump storage
plant to give an economic incentive to the optimizer to store excess energy and
not curtail it. On the other hand, if the curtailment is set too high, the optimizer
will find an economic incentive in cycling excessive renewable energy through the
conversion losses of storage plants. Herein, the curtailment cost should be capped
by the sum of the marginal cost of pumping and turbining electricity though a
pump hydro plant. The curtailment cost becomes bounded between 8.1 €/MWh
and 16.2 €/MWh. Since the O&M of PV, CSP, wind onshore and wind offshore
is different, we chose to select curtailment cost for each technology according to
the same proportional difference between them. The penalty matrix Ru per bus
becomes
Ru = diag(0, −10000, 17.6, 8.5, 22.3, 10.4, 25.5, 11.9, 34.5, 15.9, (4.10)
3.6, 0, 3.6, 0, 98.5, 0, 74.1, 0, 149.3, 0, 47.6, 0,
27.4, 0, 4.2, 0, 15.6, 0, 8.1, 8.1).
The reason we apply a different curtailment cost to each RES technology is to

push first for the integration of the renewable technology with the lowest operation
cost in case of an event where all 4 renewable infeeds are in excess.
4.3.2.3 Ramping controller

The fourth term of equation 4.5 represent the quadratic ramping cost of a load,
generation or storage node. In order to avoid an erratic behavior in dispatching
conventional power plants or curtailing load and renewable plants, penalties on
δu are introduced in the matrix δQu . The cost terms are set heuristically by PSL
to approach a realistic dispatch and to improve the numerical properties of the
dispatch optimization. They do not represent any monetary value. The entries of
matrix δQu are:
δQu = diag(0, 1e−4 , 0, 1e−4 , 0, 1e−4 , 0, 1e−4 , 0, 1e−4 , (4.11)

0, 1e−4 , 0, 1e−4 , 1e−1 , 0, 1e−1 , 0, 1e−1 , 0, 1e−1 ,
0, 1e−1 , 0, 1e−1 , 0, 0, 0)
They are set arbitrary small such that the economic dispatch that mainly deter-
mined by Ru is not distorted. According to [54], the controller does not transfer the
variability of the load profile one the shedding time series as its ramp-rate is pe-
nalized. The power generation from renewable plants is not penalized as the plants
are not controllable but curtailable. Ramp-rates of conventional power plants are
penalized to avoid erratic control actions that could could stress the mechanical
parts. Pump hydro rates are not penalized as the technology is fully flexible in one
hour4 .
4
refer to table 1.9
4.4 Controller behavior

Figures 4.2 and 4.3 illustrate the dispatch of the German power system in 2020
under a high-share of intermittent infeed, once given the priority infeed, and once
being subjected to an economic dispatch. In both scenarios, the O&M cost of
variable renewable technologies is subsidized so that their cost and curtailment
are set to zero. The scenario where variable renewable technologies compete with
conventional power is shown in A.1. We discuss the different dispatch results ob-
tained based on the different MPC controllers we have set and the strategies we
have defined previously. Please note that nuclear power plants are phased out in in
2020 therefore the ugen,S lowRamp parameter in the legend is ineffective. The ugen,CS P
parameter is as well ineffective since there will be no CSP plants installed in Ger-
many. In the legend, Ex denotes exports and Im denotes imports.
Figure 4.2: Priority dispatch of the German power system in 2020 under a high-
share renewable infeed during the month of March.
Figure 4.3: Economic optimum dispatch of the German power system in 2020
under a high-share renewable infeed during the month of March.
We explain the difference in dispatch obtained for the same German power
system under the same renewable infeeds but different dispatch policies. The main
points are highlighted:
1. Higher amounts of VRE generation are shedded under an economic dispatch

and curtailments are more frequent as can be observed when comparing the
second sub-plot of the two figures.
2. Expensive flexible plants with fast ramping capabilities are dispatched more
often in the priority infeed scenario to respond to sudden shortfall. The yel-
low curve area of the first sub-plot, which indicates the total amount of en-
ergy produced with fast ramping generators, is larger in figure 4.2 than in fig-
ure 4.3. For example, between day 15 and 16, a economic dispatch partially
curtails the high renewable infeed to ramp up inexpensive power plants and
keep them running when solar and wind energy coincide to decrease simul-
taneously during that night. A priority dispatch is constrained to integrate
fully solar and wind energy during the day and thereby turns on the more
flexible and expensive power plants during the night.
3. PSP are more frequently cycled in the priority infeed dispatch to maximize
VRE integration as is observed when comparing the third sub-plot of the
two figures. They undergo almost every day a full cycle especially when
the system is subjected to a high PV infeed on a given day. Such cycling is
ensured by the day-ahead prediction horizon of the MPC that captures the
bell-shape curve of solar infeed.
4. Since the storage state of hydro reservoirs was not penalized and the marginal
cost of hydro power is the cheapest, the reservoirs have been emptied prior
to the month of March.
5. Germany’s net balancing exchanges with its neighboring countries is not

similar under the two policies. A notable difference between the two figures
stand between day 7 to 11 which were characterized by a high PV infeed on
day 7, a high wind onshore and offshore from day 8 to 11 and a shortage of
both after. Under the priority dispatch, the German power system absorbed
all the renewable infeed and exported the rest. Day 7 exporting exchanges
are more important than the rest as there was a high solar infeed on that
day in addition to wind. On day 11, with a shortage in renewable infeed,
Germany was a net importer. Under the economic optimium, medium ramp
generators ramped up, part of renewable power was curtailed and the rest
was exported, as can be observed in the fourth subplot of figure 4.3. On
day 11, since Storage plants were filled up prior to the infeed, they were
dispatched and Germany did not import power.
6. Load shedding was kept as last resort in both scenarios and was not used.
4.5 Sensitivity analysis

The sensitivity analysis we conducted consisted of finding how the European power
system should be upgraded to absorb higher-shares of intermittent renewable en-
ergy supply and how does different renewable policies affect the power system
dispatch. Figure 4.4 illustrate how it is implemented to the simulation tool of PSL.
Figure 4.4: Structure of the MATLAB-based simulation tool with the implementa-
tion of a sensitivity analysis.
By linearly increasing the capacity of the flexibility sources existing in the

European power system of 2020, we were able to assess the benefits of all possible
combinations of power system upgrades under a high-share of intermittent supply
and filter out the optimal development based on the renewable energy integration,
the operation cost of the system and the CO2 emissions.
Part II discuss the post-processing part we performed in this Master thesis
work. First is the comparison between our MPC-controlled dispatch of the Eu-
ropean power system in 2010 with the real measurements of that year. Second is
the impact renewable policies have on the operation of the power system under a
high share RES and third is the sensitivity analysis findings.
Part II
Results and discussion
72
Chapter 5
Simulation vs Real-World
dispatch results
We verify in this chapter the validity of our power system dispatch by comparing
the dispatch measurements of some selected European countries found in the IEA
electricity report for the year 2010 [14] with our own results. We check that the
CO2 emitted from the electric power sector of each country is in line with IEA
reference values found in [19].
74
CHAPTER 5. SIMULATION VS REAL-WORLD DISPATCH RESULTS 75
5.1 Comparison of energy mixes

5.1.1 Austria
Figure 5.1: Power node dispatch in Austria vs IEA references in 2010 [14].
5.1.2 France
Figure 5.2: Power node dispatch in France vs IEA references in 2010 [14].
5.1.3 Germany
Figure 5.3: Power node dispatch in Germany vs IEA references in 2010 [14].
5.1.4 Poland
Figure 5.4: Power node dispatch in Poland vs IEA references in 2010 [14].
5.1.5 Sweden
Figure 5.5: Power node dispatch in Sweden vs IEA references in 2010 [14].
5.1.6 Switzerland
Figure 5.6: Power node dispatch in Switzerland vs IEA references in 2010 [14].
The previous figures shows that our power node dispatch has captured the sin-
gularities of every national power system and is not far from the IEA reference
values. We discuss below the similarities and differences between our dispatch for
the year 2010 and the IEA reference measurements found in [14]:
1. The contribution of hydro power to the energy mix of every country is almost
perfectly matching. This validates our method in simulating a hydro infeed
in every single country based on the hourly hydro power sery of Switzer-
land1 .
2. The shares of solar and wind energy are slightly higher in the power node
dispatch. This is due to the fact that the wind speeds used to derive the
wind power curves are not representative of the year 2009 but of 2004 as
was explained in section 2.1.2. On the other hand, it is not specified in
the IRENE-40 report [20] to which year the SODA solar infeed database
correspond to.
3. The MPC-controlled power node dispatches inexpensive controllable plants

longer than in reality. From the left-side of the figures, it can be observed
that the bulk of either coal, nuclear power and other renewable power plants2
is always higher than the bulk on the right-side of the figures which brings
the chunk of gas plants to be lower. The reason is due to the fact that we
have not set a maintenance time for the inexpensive power plants and they
are therefore dispatched as long as possible. Moreover, we have not taken
into account the different tax rates and import costs between countries and
have unified the cost of fuel in Europe. Thereby, it would be impossible to
obtain an exact match between our country dispatch and reality. Important
is to capture the trends of generation mixes in each EU country.
Table 5.1 and Table 5.2 displays the power balance reference values from [14]
and the ones obtained with the power node dispatch for the previously illustrated
energy mixes.
1
refer to section 2.3.2
2
consist of biomass and geothermal plants
Table 5.1: Power node dispatch vs IEA reference values in 2010.
Austria France Germany

(GWh) Power Node IEA Power Node IEA Power Node IEA
Generation 65977 70700 602970 544600 580120 592400
Consumption 64778 62700 505480 471800 554846 543800
Storage 6863 4600 10125 6600 12682 8600
Import 18521 19900 14002 19400 53990 43000
Export 12857 17600 101370 50200 66580 57900
TR3 losses - 3400 - 35410 - 24001
Power Balance4 0 0 0 0 0 0
Table 5.2: Power node dispatch vs IEA reference values in 2010.
Poland Sweden Switzerland

(GWh) Power Node IEA Power Node IEA Power Node IEA
Generation 138104 142900 136006 145300 67211 66100
Consumption 142084 128800 127404 135000 63971 59800
Storage 0 0 17 100 1624 2500
Import 11866 6300 26356 14900 35387 33400
Export 7885 7700 34942 12900 37004 32900
TR losses - 10302 - 10601 - 4404
Power Balance 0 0 0 0 0 0
Surely the values cannot match exactly and the power exchanges are different
due to the previously explained reasons of point 2 and 3. Additionally, the load
series provided by IRENE-40 correspond to the ENTSO-E data of year 2009 is no
longer published online. However, important to note is that the order of magnitude
of the generation, consumption and storage values is similar and that the power
balance is respected which validates our model and dispatch.
5.2 CO2 emissions

The CO2 emissions is a significant parameter in the post-processing part of our dis-
patch and has to be conformant to reality so that we can estimate feasible emissions
reductions. Based on [19], we assign a ton of CO2 emitted per MWh of electricity
produced for each type of power plant5 . Figure 5.7 illustrates the CO2 emissions
3
TR = transmission
4
Power Balance = Generation - Consumption - Storage + Import - Export - Losses
5
0.92 tonCO2 /MWh-coal, 0.62 ton CO2 /MWh-Oil, 0.37 CO2 /MWh-gas
emitted from the power sector of each EU country in 2009 according to [19] and
the ones obtained by dispatching the 2010 system using the power node method.
Figure 5.7: CO2 emissions comparison: Power Node vs IEA reference.
We notice that the CO2 emitted per country using the Power node dispatch
(red curve) is in line with the IEA reference emissions (black curve) found in [19].
The total emissions from the EU power sector results to 1.3 billion tons in 2010
according to the IEA and 1.05 billion tons according to power node dispatch in that
same year. The latter did capture the highest polluting EU countries (Germany,
United Kingdom and Poland) as in reality.
Based on the results shown in this chapter, we have validated our method in
modeling the European power system and using an MPC-controlled optimizer to
dispatch. The next chapter presents the analysis we have conducted to study the
different impacts a high shares of fluctuating renewable energy sources will have
on the system.
Chapter 6
The impacts of a high-share of

renewable energy supply on the
European power system
We expose in this chapter the outcome of prioritizing the RES infeed on the opera-
tion of the European power system. We compare such dispatch to an economic one
where the priority infeed is removed, thereby allowing for economic curtailments
of var-RES. The results demonstrates that under a scenario where variable renew-
able energy covers 53% of Europe’s electricity demand, it will be more optimal to
allow for economic curtailments than just balancing curtailments. We present also
the changes in power flow patterns around France under a high-renewable supply
and prove that the lines will be less loaded on average in a high-RES scenario than
in a low-RES scenario. We assess the impact the former will have on the electricity
generation cost in Europe and the market price. If the RES shares are increased
from 12% to 53%, the generation cost of electricity will likely drop by 82% and
the market price by 27%.
82
CHAPTER 6. THE IMPACTS OF A HIGH-SHARE OF RES ON THE EPS 83
6.1 The impact of renewable policies

6.1.1 Scope of analysis
The European power system (EPS) operation cost is found using the equation be-
low:
T X
EU X
G
X e
EPSOp (e) = MCg ( ).Eg(i,t) (MWh) (6.1)
t=1 i=1 g=1
MWh
The first term sums up the total cost of operation (including pump storage
plants) of a single country at the first hour. The cost of operation is found by mul-
tiplying the marginal cost (MC) in 2020 of every type of node defined in table 2.7
with the electricity generated (Eg ) from the respective node at this hour. The cost
of operation is then summed up for all Europe and the hour index is incremented
by one step. The loop repeats itself till the hour index reaches 8760.
To find out how renewable policies affect the operation of the system in a high-
share RES scenario, we fixed some variables for this study:
• The EPS analyzed is based on all the generation, transmission and storage
capacities of 2020.
• The load demand used correspond to the 2050 load forecast provided by
IRENE-40 (total of 5050 TWh).
• The marginal cost of generation and storage used is the one assigned in table
2.7 for 2020.
• A low RES availability scenario is defined corresponding to the wind and

solar infeed in 2020 (12% of 2050 load demand [20]).
• A high RES availability scenario is defined corresponding to the wind and

solar infeed in 2050 (53% of 2050 load demand [20]).
To evaluate the impact of the priority infeed policy and an economic dispatch,
once with a subsidized O&M cost for varibale renewable plants and once not, we
varied the input terms of the MPC-controller according to equations 4.8, 4.9 and
4.10. The post-processing of the three different policies applied to a low and high
renewable scenario is based on analyzing the integration of renewable energies to
the grid, the CO2 emissions and the operation cost of the system as defined in
equation 6.1.
6.1.2 Dispatch results

The table below illustrates the results obtained in dispatching the same power sys-
tem under different renewable infeeds and different policies
Table 6.1: Comparison of dispatch policies.
Priority infeed Economic dispatch

O&M subsidized O&M subsidized RES compete
RES availability 2020 2050 2020 2050 2020 2050
RES integration 100% 93% 98.7% 84% 93.9% 74%
CO2 em.(Mn ton) 1684 834 1688 873 1694 936
Op. cost (bn €) 183.42 97.12 183.69 88.85 199.47 141.75
Valuable information can be concluded from these results:

1. Under the wind and solar energy availability in 2020 (500 TWh), the power
system is able to absorb the full amount of variable renewable energy (VRE)
when it is given priority infeed. Under an economic dispatch, 6.5 TWh
(1.3%) is curtailed if the O&M cost of variable renewable plants (VRP) is
subsidized, 30.5 TWh (6.1%) is curtailed if VRP have to pay for their pro-
duction as cheaper controllable plants (geothermal, hydro, pump storage) are
dispatched more often.
2. Under the wind and solar energy availability in 2050 (2500 TWh), the power
system reaches a limit in absorbing VRE at 2325 TWh (93%), when it is
given priority infeed. This highlights the technical limit of the 2020 EPS in
integrating high-shares of VRE as the sources of operational flexibility1 have
been used to their extent.
3. The CO2 emitted from the European power sector is directly related to the
renewable energy integration. The more RES is integrated, less CO2 is emit-
ted. Compared to the 2010 emission, the best scenario (high renewable in-
feed given priority) reaches only reduction of 21%, emphasizing that decar-
bonizing the power sector does not solely depend on having more renewables
integrated but a change in the 2020 EPS fleet is needed.
4. In 2020, the operation cost of the system was lowest when RES integration
was highest, i.e under the priority infeed. More free energy was fed in to the
grid which results in benefits of €Mn 270 compared to when its curtailment
was allowed. Of course when the O&M is no longer subsidized, the opera-
tion cost of the system is €bn 15.78 higher. This subsidy can represent the
emission abatement cost of 6 million tons of CO2 in the atmosphere.
1
refer to figure 1.6
5. In 2050, if the O&M cost of VRP was subsidized, operating the EPS under a
priority infeed costed €bn 8.27 more than under an economic dispatch. This
quantifies the value of curtailing the energy output of wind and solar plants
when their shares are high. The reason behind such difference is further
illustrated in table 6.2 and explained.
6. When moving from a low to a high renewable energy scenario, the savings in
operation cost per year varies between €bn 57 to 95 depending which policy
is chosen. These savings represent mainly fuel cost savings for Europe. They
are significantly important due to the economy of scale effect.
The table below displays the aggregation of the dispatched units of the 2020
EPS under a a high-renewable infeed but subject to the priority infeed policy and
the economic dispatch where the O&M of variable renewable plants is subsidized
in both cases. The difference in operation cost between the two dispatch sums up
to €bn 8.269.
Table 6.2: Priority infeed vs economic dispatch.
Plant type Priority infeed Economic dispatch

Conventional energies Yearly (TWh) Yearly (TWh) Difference MC (€/MWh) Operation (Mn €)
Gas 254.35 120.29 134.05 84.54 11332.36
Coal 604.71 787.14 -182.42 47.45 -8655.29
CCGT 276.45 228.92 47.63 74.08 3528.52
Oil 48.37 7.66 40.71 149.26 6076.39
Nuclear 640.97 931.54 -290.57 15.65 -4547.06
Non-variable RES
Hydro RoR 139.25 119.39 19.86 3.86 76.69
Hydro Res 384.59 384.58 0.01 3.86 0.03
Geothermal 9.57 13.15 -3.58 4.21 -15.04
Biomass 219.36 221.42 -2.05 24.79 -50.91
Storage plant
Turbine 96.68 67.05 29.63 8.12 240.49
Pump 112.39 77.52 34.86 8.12 282.93
Variable RES
PV 924.19 839.96 84.23 0 0
CSP 115.75 101.05 14.7 0 0
Wind onshore 726.67 655.12 71.55 0 0
Wind offshore 724.05 653.04 71.05 0 0
Total 8269.1
One directly notices that under the priority infeed, higher amounts (170.5 TWh)
of variable RES are integrated compared to an economic dispatch. However, since
the economic curtailment of renewable energy is not allowed to turn on low flexible
plants prior to a predicted shortfall, highly flexible plants are used more often that
in an economic dispatch to balance sudden renewable energy shortage. 222 TWh
of Gas, CCGT and Oil has been dispatched which costed additionally €bn 20.94.
Also, as we had highlighted in figure 4.22 , storage plants are effectively more cy-
cled to integrate renewable energy and this comes for an extra cost of €mn 523.42.
Likewise, less variable renewable energy is integrated under an economic dispatch
and 479 TWh of low-flexible generation (Coal, Nuclear, Biomass, Geothermal) can
be dispatched which leads to reductions in operation of €bn 13.27. The final value
obtained proves that it will be necessary to shift policies in the future as economic
curtailments can help save €bn 8.27 per year when operating the European power
system under high-shares of RES.
On the other hand, it would be misleading to conclude that in an economic
dispatch where VRPs benefit from a subsidy, PV energy is most curtailed and CSP
energy is most integrated as the difference in energy generated between the two
policies is highest for PV plants and lowest for CSP plants. One need to compare
the ratio of renewable energy integration to availability. The next section elaborate
on this issue.
6.1.3 Renewable energy integration

We investigate the impact different renewable policies have on renewable energy
integration to the grid. Table 6.2 indicate the sensitivity of the integration of each
type of renewable infeed to the renewable policy chosen. Here, both policies ben-
efit from a subsidy that nullifies the marginal cost of production of each type.
Thereby, there integration is purely dependent on the shape of the infeed curve3 .
Table 6.3: The influence of policies on the integration of each type of renewable
energy infeed.
PV CSP Wind onshore Wind offshore

TWh Priority Economic Priority Economic Priority Economic Priority Economic
generation 924.19 839.96 115.75 101.05 726.67 655.12 724.05 653.04
curtailment 46.34 130.6 14.06 28.76 71.05 142.6 56.89 127.9
available 970.53 970.53 129.81 129.81 797.72 797.72 780.87 780.87
Int. (%) 95.2 86.5 89.2 77.8 91.1 82.1 92.7 83.6
From the table we can read that the optimizer integrated most PV energy, a
similar amount of wind onshore or offshore and lastly CSP energy. Of course, in
2
This figure illustrated the controller behavior under a priority dispatch
3
refer to figures 2.4 and 2.5
the priority infeed policy, the ratio of integration is higher as the curtailment cost
was set extremely high to push for the integration of renewable energies and allow
only for emergency curtailments and no economic one. The reason behind such
numbers can not be explained as it depends on the path the optimizer chose to
follow to find an optimal solution. We reckon that the predictable bell-curve shape
of PV infeed could allow for a smooth cycling of storage plants, thereby PV is the
less curtailed RES. Since the CSP infeed is constant during one day, its curtailments
provide an easier balancing solution whereas the wind onshore and offshore infeed
is variable, thereby the balancing solution becomes more challenging.
The last policy we analyze is the pure economic dispatch where the subsidy for
VRPs is removed and they must compete against conventional plants. In that case,
we had set a curtailment cost according to the least expensive O&M cost of VRPs
and in between the cycling cost of storage plants 4 . Figure 6.1 illustrates the energy
integrated of each type of plant according to the policy chosen.
Figure 6.1: Influence of renewable policies on renewable integration results.
The first two blocks resonate the values shown in table 6.2. The third block
shows a different dispatch. Because the cost of curtailment of wind offshore plants
was set the highest in this scenario, the controller indeed shedded the most out of
this variable supply. Since PV plants have to pay €17.6 per MWh of electricity
produced, the optimizer integrates by 1.5% more PV energy in a pure economic
dispatch as in a subsidized scenario. Even though the curtailment cost for wind
4
refer to section 4.3.2 of chapter 4
onshore was set €1.5 higher than of CSP, optimizer still integrated more wind
onshore energy.
The renewable policy chosen definitely influence the dispatch outcome. The
next section has been simulated under the assumption that the priority infeed to
renewables will be removed but that the O&M cost of renewable plants will still be
subsidized.
6.2 The impact on power exchange patterns

Traditionally, power systems have been operated such that supply follows fluctu-
ations in demand. Extrapolating this trend to country exchanges, exporting (sup-
plier) and importing (consumer) countries in Europe have always been known.
With the largest capacity of nuclear power plants in Europe, France have always
been a net exporter of power. Only when air temperatures drops very low and the
heating demand peaks, France is a net importer of power as most of the houses are
heated withNet$balancing$exchanges$in$France$in$2010$'ENTSOCE$
electric heaters. The 2010 ENTSO-E data for France proves this point.
values'$
7000$
6000$
5000$
4000$
MW$ 3000$ Net$balance$(Ex!Im)$
2000$
1000$
0$
Jan$ Feb$ Mar$ Apr$ May$ Jun$ Jul$ Aug$ Sep$ Oct$ Nov$ Dec$
!1000$
Figure 6.2: Net balancing exchanges between France and its neighbors in 2010
according to the ENTSO-E values [15].
In that year, the month of January was very cold which explains why France
3000$
was a net importer from its neighbors (Belgium, Germany, Italy, Spain, Switzer-
land, United Kingdom).
2500$
Given this trend, we analyzed according to our power node dispatch how the
lines would be loaded around France in a low and high-share renewable supply;
low being the RES availability in 2020 and high the RES availability in 2050 ac-
2000$
cording to the provided IRENE-40 data. Figure 6.3 illustrate the power exchanges
1500$ BE$
CH$
DE$
MW$ 1000$
ES$
GB$
500$ IT$
0$
Jan$ Feb$ Mar$ Apr$ May$ Jun$ Jul$ Aug$ Sep$ Oct$ Nov$ Dec$
of France with its neighbors under the availability of RES in 2020. The line loads
are found by averaging the hourly imported or exported power on a weekly basis
and is expressed as % values of the planned NTC in 2020. Table 6.4 displays the
expansions of the NTC lines of France with its neighbors according to [61].
Figure 6.3: Line loading in 2020 around France in a low-share renewable energy
supply scenario.
Table 6.4: NTC installed in the European power system of 2020
From To Export (MW) Import (MW)

FR UK 3000 -3000
FR BE 3375 -2025
FR DE 2700 -3125
FR ES 4250 -3500
FR CH 3100 -1700
FR IT 4225 -2633
The high density of red squares from the spring to fall period that appears on
the upper half of figure 6.3 and dark blue squares on the bottom can be noticed.
This indicates that the exporting lines were loaded closed to their NTC limit and
the importing lines not more than 10%. This proves that France was a net exporter
during this period. On the other hand, during the winter, we notice different line
loading patterns due to the high demand of electric heaters in France. Very inter-
esting is the first week of exchanges between France and its neighbors where lines
are almost fully loaded in the direction of imports. This is explained by the strong
ramping constrained applied to nuclear power plants which need time to reach full
capacity. From this picture two conclusions can be drawn. First, under the RES
availability of 2020, the power flow patterns did not change compared to today.
Second, in case TSOs are planning to expand power line capacities, priority should
be given to the exporting lines from France as these were loaded close to their limit
for a long period of time. However, with more intermittent supply in the European
energy mix, power flow patterns will change and the conventional way of planning
line expansion is no longer valid. Figure 6.4 illustrates this point.
Figure 6.4: Line loading in 2020 around France in a high-share renewable energy
supply scenario.
Our first observation is that the exporting lines of France are no longer fully
loaded from the spring to the fall period, whereas the importing lines are more
often loaded between 20% and 40%. This applies specifically to Italy which is
exporting PV energy during summer weeks whereas in a low-renewable scenario
the line was merely loaded. Only the power exchanges with Switzerland follows
the same pattern but this is due to the fact that the share of RES in Switzerland are
below 1 % according to the IRENE-40 scenario. Second observation is that the
total number of red squares have decreased in a high-renewable scenario compared
to a low one which indicates that the system is more secure since the lines are no
longer loaded at 100% of their NTC as much as before.
The main message from this section is that the transmission network should
not be expanded in the traditional way to host more renewable energy. It does not
mean that if specific lines have been highly loaded today that they will be so in the
future if more variable renewable supply is fed in. By conventionally expanding
the transmission capacity, TSOs could over-evaluate the needed transfer capacity
in this region and under-evaluate the one in another region. Also, the common
belief that if more var-RES are fed in, then the power system lines will be more
heavily loaded is not an absolute truth. We showed in figure 6.4 that the power lines
connecting France and its neighbors in 2020 were less loaded, on a weekly average,
under the 2050 var-RES availability than the 2020 one, given that the former is five
times higher than the latter. This does not mean that the power line NTCs should
not be further expanded to boost RES integration. It just means that the western
European power system security is less threatened. In appendix B, the reader can
find the same plots for the line loadings in Northern, Central and Eastern Europe
under a low and high RES supply, which illustrate the same observation. How to
optimally expand each transmission line in Europe falls outside of the scope of this
thesis. We present however in the next chapter how we find the limit to expand the
total transmission capacity needed to maximize renewable energy integration and
minimize the system cost.
6.3 The impact on power markets

6.3.1 The electricity market price forecast
The market price of electricity is found on quarterly basis when the submitted
day-ahead bid and demand curves intersects [43]. Basically, the last dispatched
power plant that has produced 1 MWh of electricity defines the price of electricity
at this moment in €/MWh according to his willingness to produce this last unit
of electricity. Based on this rule, the market price in every EU country can be
plotted according to different renewable supply and policies, different power sys-
tems, different load demands and different marginal cost of production. We fixed
in this analysis the power system and marginal cost of production to 2020, the load
demand to 2050 and varied the renewable supply to study how the market price
will change. Moreover, the priority infeed is removed and renewables plants must
compete in the market. Figure 6.5 displays the hourly spot price of electricity in
Portugal during the summer of each year based on such configurations.
Spot prices in PT in 2020

150
125
Eur/MWh
100
75
50
25
0
June July August September October
150
125
Eur/MWh
100
75
50
25
0
150
125
Eur/MWh
100
75
50
25
0
150
125
Eur/MWh
100
75
50
25
0 X: 209 X: 828 X: 1841 X: 2455
June Y: 22.37 July Y: 24.79 August Y: 17.66
September Y: 25.51 October
Figure 6.5: Forecast of the market price of electricity in Portugal.
From the evolution we notice first that the variability of the spot price increases
with the increase of RES availability. In 2050, the price jumps from 17 to 150∼
€/MWh much more frequently that the scenarios with lower shares of RES. Sec-
ond, we notice that the frequency of the spread between base (84 €/MWh5 ) and
peak price (150 €/MWh6 ) observed in 2020 decreases as RES availability increase
till 2040. On that year, oil power plants were dispatched only twice from June to
October. However, in 2050, oil power plants were dispatched more than twice but
the price of electricity has settled many times below 26€/MWh. The four boxes
added to the last sub-plot indicates the marginal cost of production in 2020 for
each type of variable renewable plant we have defined in table 2.7. This indicates
that the demand was met more often with just variable renewable power but that
oil power plants were more often needed for fast ramping balancing requirements.
Overall, figure 6.5 finally shows us that the electricity price in Portugal on average
5
Gas power plant
6
Oil power plant
will decrease with more renewables in the country energy mix. The latter was at
97.56 €/MWh in 2020, 87.25 €/MWh in 2030, 78.32 €/MWh in 2040 and 67.11
€/MWh in 2050. When increasing the share of RES in Portugal by 10%, 18%, 35%
compared to the RES availability share in 20207 , the electricity price decreased by
11%, 20%, 32%. The decrease in spot price of each European country is shown in
the appendix in B.8.
6.3.2 The electricity generation cost

The electricity generation cost in each European country i is found using the equa-
tion below:
PT PG e
e t=1 g=1 MC g ( MWh ) · E g(i,t) (MWh)
egen,i ( )= PT PG (6.2)
MWh t=1 g=1 E g(i,t) (MWh)
We take the weighted average of the latter to express the EU electricity gener-
ation cost using:
P29 e
i=1 egen,i ( MWh ) · E g(i) (MWh)
ēgen,EU = P29 (6.3)
i=1 E g(i) (MWh)
By fixing the same configuration as the previous subsection and varying the
renewable supply from low to high-shares in a scenario where VRPs have to com-
pete with conventional plants, we plot in figure 6.6 the forecast of the electricity
generation cost in the EU and the weighted average spot market price.
7
18% of the 2050 load demand in Portugal, 91 TWh [20]
Figure 6.6: Forecast of the electricity generation cost and market price in Europe
when the O&M cost of VRPs is considered.
When increasing the share of RES by 7%, 19% and 41% compared to the
RES availability share in 2020 in Europe8 , the electricity generation cost in Europe
dropped by 53%, 72%, 82% and the market price of electricity dropped by 9%,
19%, 27%, even when not considering renewable energy supply as a free source
of power but taking into account the O&M cost of such plants. Figure 6.6 clearly
emphasizes the economical benefits of a non-subsidized high-renewable supply.
The macro-economics benefits of such reductions in price and generation cost falls
outside of our scope of study. The results can be used for further studies.
We explained in this chapter the different technological and economical im-
pacts a high-share renewable energy supply have on the European power system of
2020 and specially the impact renewable policies can have on the operation of the
system and the integration of renewable energies. The main goal is to now develop
the system such that we can maximize the renewable energy integration and min-
imize the operation cost of the system while making sure that the CO2 emission
target of 2050 will be reached.
8
12% of the 2050 load demand in Europe, 5050 TWh [20]
Chapter 7
Sensitivity analysis results
This chapter presents the core results of our work; how to value the flexibility of the
power system for the integration of high-shares of renewable energy supply (RES).
We present first the results we obtained for the European power system (EPS). We
discuss the sensitivity of RES integration to an increase in capacity of the flexibility
sources in the EPS and evaluate the operational benefits that each upgrade brings to
the system. The latter represent the value of flexibility of the source in question. We
discuss also the RES sensitivity integration in Germany and Spain, the two leading
investors in renewable energies, which is unique to each country and different than
the general sensitivity of the EPS. We value the flexibility sources in each country.
Finally, CO2 emissions in the EPS are analyzed with an increased flexibility of the
system. Modifications to the power system fleet are mandatory, otherwise the 2050
CO2 emissions reduction target can not be reached. Please note, in all different
power systems setups were adequate. No load shedding occured above 0.2% in
each EU country.
96
CHAPTER 7. SENSITIVITY ANALYSIS RESULTS 97
7.1 EU level
To perform our analysis, we fixed first the load demand in the EU and the var-RES
infeed to the 2050 forecast provided by IRENE-40 project and started our simula-
tions with the European power system (EPS) configuration of 2020. We do so to
evaluate the flexibility of the latter and measure how increasing it affect the oper-
ation of the system. All scenarios simulated corresponds to an economic dispatch
where the O&M cost of VRPs is assumed to still be subsidized1 . As we have men-
tioned in the introduction, the additional flexibility coming from demand response
(electric vehicles or thermal loads) is not evaluated due to time constraints. We
focus on the added operational flexibility an expansion of the power line transmis-
sion capacities and storage (power or energy rating) capacities brings to the system
as well as the situation where controllable generators have faster ramping rates.
7.1.1 RES integration results

7.1.1.1 NTC vs storage power rating
Figure 7.1: RES integration sensitivity relative to the power lines NTC and storage
power rating expansions in Europe.
1
By 2050 the O&M cost of VRPs will be lower due to learning curve so the subsidy is small
First to notice is the bottom RES integration value of 83.95% which represents
the flexibility measure of the 2020 EPS under the 2050 RES availability and cor-
responds to the value shown in table 6.1 of the previous chapter. The figure then
displays the sensitivity of RES integration in the power system to the linear capac-
ity increase of power lines NTC and storage power rating capaities. For example,
we read that doubling the storage power rating in Europe to 94 GW helps boost the
integration of 43 TWh of intermittent supply (1.6% of 2675 TWh) and that dou-
bling the total NTC capacities installed to 340 GW can help boost the integration of
115 TWh (5% of 2675 TWh), while doubling both flexibility parameters together
can help integrate 20 TWh more than the latter. With a higher pump storage power
rating (SP ), it would be possible to charge and discharge the pump storage reser-
voirs more rapidly, thereby allowing to store higher excesses of renewable energy,
while increasing the NTC between the EU country increases the interconnected-
ness of the system and allows each area to make use of the other area’s flexibility
sources. The less TSO’s curtails RES, the higher the fuel cost savings area. 1%
of curtailed RES from a demand of 5050 TWh in 2050 translates in operational
cost savings of €bn 0.79 to 4.3, if this energy was to be generated with nuclear
power or gas power plants with fuel and CO2 prices of 2020. Table 7.1 depicts the
exact reductions in operation cost of the system when the flexibility of the system
is increased.
Table 7.1: Operation cost of system update (bn € per year).

PP
PP SP
P 1 2 3 4 5
NTC PP
P
P
1 88.92 87.1 86.2 85.8 85.6
2 79.1 77.4 76.7 76.4 76.3
3 73.4 72.1 71.4 71.1 70.9
4 70.3 69.1 68.5 68.2 68.1
5 68.9 67.7 67.1 66.9 66.7
Table 7.1 highlights the operational benefits in investing in these two flexibility
sources. It falls outside of the scope of our study to perform a detailed investment
analysis. Roughly calculated, with a capital cost of 1 €bn /GW for power lines
[1] and a capital cost of 0.12 €bn /GW for the equipment of pump storage plants3 ,
doubling the capacity of both sources will come at a cost of €bn 175 which can
break even in 15 years, as the operation of the system cost €bn 11.5 less per year.
Please note that the capital cost for power lines have been averaged according to the
costs listed in table D.1 and table D.2 for planned projects of 1000 MW capacity
2
Same value as listed in table 6.1 when the O&M cost of VRPs is subsidized.
3
The equipment cost consist of the cost of the pump, turbine, generator, transformer. The equip-
ment cost ration has been found according to [66] and applied to the total cost of PSPs listed in
[64]
in Europe and the operational benefits calculated are based on the cost of fuel and
CO2 emissions in 2020, as well they have not been discounted.
7.1.1.2 Storage power vs storage energy
Figure 7.2: RES integration sensitivity relative to storage power and storage energy
expansions in Europe.
Figure 7.2 illustrate the different sensitivity of RES integration to the storage power
(SP ) and energy rating (SE ) expansions in Europe. The integration is more sensitive
to the power rating capacity than the energy reservoir capacity, meaning to the rate
at which we can store excess energy and not the available quantity of energy which
we could store in the reservoir. However, this figure correspond to the aggregation
of all power and energy rating of pump storage plants in Europe and interpreting
the results in this way can be misleading. It is crucial to look at country to country
results to figure out how does the increase of storage energy or power rating influ-
ence the RES integration. The next section will elaborate on this issue for Germany
and Spain, the two leaders in renewable energy investments. From a system-level
perspective, if all countries simultaneously increase the power and energy rating of
their PSPs, then the European power system operation cost gets further reduced as
illustrates the values in table 7.2.
Table 7.2: Operation cost of system update (bn € per year).

HH S
H E 1 2 3 4 5
SP HH
H
1 88.85 88.42 88.19 88.02 87.88
2 87.06 86.32 85.90 85.66 85.48
3 86.24 85.33 84.83 84.49 84.25
4 85.80 84.79 84.25 83.88 83.59
5 85.55 84.49 83.89 83.51 83.22
7.1.1.3 Ramping rates

We varied the ramp rate constraints applied to the fleet of conventional power plants
according to table 2.3 of Chapter 2. Depending on the TSO choice to keep a portion
of specific types of power plants turned on, the rate to ramp up these plants from
start will differ, and thereby, the economic dispatch of the power system as well. If
the capability of ramping nuclear power plants is higher, then these conventional
controllable plants will be dipstached more often since they have the least marginal
cost of production.
We found out that if we kept the quadratic cost of ramping rates defined as
in section 4.3.2.3, changing the ramping constraints was ineffective. The RES
integration and the operation cost of any simulated power system was still fixed at
84% integration and €bn 88.9. The values only varied by range of 0.03% between
each other. The explanation is that the quadratic cost used allows the optimizer
to recognize that abrasive ramping will damage the plant, therefore no matter how
much we increase the ramping capacity of a plant, the process of ramping a plant
has a high (quadratic) cost and therefore the dispatch will not significantly change.
We modified the quadratic cost ramping controller vector by removing the con-
straint applied to conventional generators4 and allowing for abrasive operations.
The results we obtained when varying the ramping constraint of the 2020 EPS
under the var-RES availability in 2050 are shown in table 7.3
Table 7.3: Flexibility measure for the ramping rates of conventional generators.
State of fleet RES integration Op. Cost

All fleet turned off 91.6 % 76.2 bn €
75% of fleet turned off + 0.35 % -0.7 %
All fleet turned on + 1.62 % -2.6 %
4
1e-1 term appearing the δQu controller vector
The first observation is that the RES integration value of the 2020 EPS un-
der a high-renewable infeed jumped from 84% to 91.6% and the total operation
cost of this system decreased from €bn 88.9 to €bn 72.2. The reason is that by
allowing for excessive bi-directional ramping of conventional power plants, RES
are curtailed less and therefore the operation cost of the system decreases5 . The
more we increase the rampable capacity of the power plants given that there is
no abrasive cost assigned, the higher RES integration and lower the operation cost.
Table 7.3 shows that in the best scenario, where we applied no ramping constraints,
the flexibility measure of the EPS can increase by a maximum of 1.62% from the
worst case scenario and the savings in operation reaches €bn 1.98. However, such
scenario does not reflect reality as conventional power plants will never be dis-
patched in this way. We therefore concluded that to analyze the ramping influence
on the integration of RES, it is necessary to run a 15 minute dispatch and constrain
each generator in the power system to a minimum load operation and its start-up
time. The optimization formulation of such problem results in a mixed-integer
programming problem. The computational time of a European power system with
dis-aggregated power plants would be extremely slow. We reckon isolated country
systems should only be studied to evaluate the ramping influence.
7.1.2 Synthesis graph

The next graph presents the sensitivity of RES integration when combining the
effect of increasing all flexibility parameters together. We track back the flexibility
measure of the 2010 European power system.
5
Figure A.2 and figure A.3 illustrates the dispatch of such explanation
Figure 7.3: RES integration vs increase of each flexibility parameter and their
combination.
First to notice is the starting RES integration value (84%) which corresponds
to the flexibility measure of the 2020 European power system (EPS) under the
var-RES infeed of 2050. We simulated the EPS configuration of 2010 under a
high-shares of RES and found that the flexibility measure reached a limit of 76%.
Therefore, if the NREAP of 2020 and the 10 year development of the ENTSO-E
are to be respected, the EPS gains already an 8% increase in its flexibility measure.
If then, the operational flexibility sources of the 2020 EPS are developed, this
graph synthesis the flexibility grades of each power system upgrade. We observe
that RES integration is most sensitive to increasing the net transmission capacity
between the countries with an almost linear slope of two percentage points, while
for storage power rating the slope of integration is one percentage point and for
storage energy rating the slope of integration is 0.5 % point. If the parameters are
increased all at once, doubling them adds 5% to the flexibility grade of the EPS,
tripling them adds 10 %, quadrupling them adds 12% and quintupling them adds
13.5%. From these results, we notice that there exist a ceiling to the integration
of such high-shares of RES (2675 TWh6 ). After tripling the flexibility sources,
the rate of increase of the flexibility grade starts to diminish and RES integration
6
53% of the European load demand in 2050
saturates. We can conclude from such graph that it would be unnecessary to have a
total NTC capacity in Europe that exceeds 510 GW, a power rating storage capacity
that exceeds 141 GW and an energy one that exceeds 7980 GWh as the operational
benefits will start to saturate and the investment in such over-capacities will be
wasteful. The next section illustrate this point.
7.1.3 Value of Flexibility in the European power system

We value the expansion of each flexibility source according to the operational ben-
efits it brings to the power system. The value can be found by subtracting the initial
operating cost of the 2020 EPS under the 2050 RES availability (€bn 88.9) by the
cost calculated for each upgrade. Table 7.4 presents these results.
Table 7.4: The value of flexibility (€bn per year) of the European power system.
Parameter NTC Storage Power Storage Energy EPS Flex

Cap. of 2020 170 GW 47 GW 2660 GWh -
Double 9.8 1.8 0.43 12.04
Triple 15.5 2.7 0.66 18.42
Quadruple 18.6 3.1 0.83 21.58
Quituple 20 3.3 0.97 23.08
Table 7.4 makes it clear that the European power system flexibility value sat-
urates as doubling first all flexibility parameters results in €bn 12.04 of savings
per year of operation, tripling it adds €bn 6.38, quadrupling it adds €bn 3.16 and
quintupling it adds €bn 1.5. Roughly calculated, if we consider this time the full
cost of pump storage plants (2.4 €/GW), which takes into account the expansion
of the energy reservoir, the investment cost for doubling the capacity of the three
flexibility sources in Europe becomes €bn 282.4 and will pay back in about 25
years; tripling the capacity will need 50 years to pay back7 , quadrupling 100 years
quintupling 200 years. Obviously such investments are wasteful and there is no
economic sense in pushing for 100% integration of variable renewable supply.
The values obtained above represent a European system-level perspective. Na-
tional TSOs, energy utilities and investors would be much more interested in know-
ing the unique flexibility value of each country’s power system. We present in the
next section the RES sensitivity integration results and the flexiblity value of Ger-
many and Spain, the two countries who are forecasted to have, each, 10% of the
total RES availability in Europe in 2050 in their own country.
7
savings are half the savings when doubling the capacity
7.2 Country level: Germany & Spain

The capacity of variable renewable plants in both countries is forecasted by IRENE-
40 to increase in 2050 according to the following:
Table 7.5: Capacity forecast (in GW) in 2050 of solar and wind power plants in
Germany and Spain [20].
Germany Spain
Cap. in year PV CSP Wind On. Wind Off. PV CSP Wind On. Wind Off.
2011 25 0 27 0.15 4 0.6 20 0
2050 160 0 78 55 109 19 76 16
With such an increase in capacity, the RES availability according to the SODA
solar radiation in Europe database and the TradeWind wind profile series reaches
495 TWh in Germany and 523 TWh in Spain, when it was at 72 TWh and 48
TWh in 2010 in both countries. It will definitely matter for both countries to min-
imize the curtailment of var-RES under an economic dispatch, as the latter has
been proven to be more economical for the operation of the system compared to a
priority dispatch. In that perspective, the operational flexibility of each power sys-
tem has to be developed according to the existing configuration and the sensitivity
of RES integration to it has to be investigated. We present here our analysis and
proposed solutions.
7.2.1 RES integration results

7.2.1.1 NTC vs storage power
Figure 7.4: RES integration sensitivity relative to power lines NTC and storage
power expansions in Spain and Germany.
The two selected countries show a much different RES integration sensitivity re-
sults than the European system-level results illustrated in figure 7.1 for the expan-
sion of NTC and storage power rating. The RES integration in the 2020 German
power system (GPS) was insensitive to the expansion of the power rating of its
pumped storage plants while very much sensitive to the expansion of the power
line NTCs connecting Germany to its neighbors. On the other hand, the RES in-
tegration in the 2020 Spanish power system (SPS) was as sensitive to the increase
of each parameter. Without increasing the capacity of its flexibility sources, the
2020 GPS able to absorb 416 TWh of the available 495 TWh (84%) of intermit-
tent supply while the 2020 SPS absorbed 424 TWh of the 523 TWh (81%). If the
power lines NTCs connecting Germany to its neighbors were doubled in capacity,
then the GPS can absorb an extra 30 TWh of renewable energy, which represents
almost half of what is integrated in the GPS today. If the storage power rating of
pumped hydro plants in Spain is doubled, an extra 22 TWh of clean and free sup-
ply of energy can be absorbed by the Spanish power system, which also represent
almost half of what is integrated in the SPS today. If such flexiblity sources are not
expanded, then each power system will have to curtail such amounts of energy and
use conventional fuel to meet the electricity demand, which in turn, will results in
higher operating cost. We present further in this section the operational benefits
such flexibility expansion brings to each power system.
It is important to note that these figures represent the aggregated results of the
four different renewable supply integrated to each grid. In the case of Germany,
PV and wind energy(onshore and offshore) exhibited the same sensitivity trend,
however in Spain, PV and CSP energy were more sensitive to the expansion of the
pumped storage plants power rating than the expansion of the power line NTCs,
while for wind energy it was the inverse situation. Figure C.1 and figure C.2 illus-
trate this point in the appendix.
7.2.1.2 Storage power vs storage energy

The sensitivity of RES integration to the expansion of the third flexibility source in
each country differed quite well compared to the one on a European system-level
perspective (refer to figure 7.2). The figures below illustrate this difference.
Figure 7.5: RES integration sensitivity relative to storage power and storage energy
expansions in Spain and Germany.
If the RES integration in Germany was insensitive to increasing the rate at

which excess energy can be stored in reservoirs, the reason was that the volume of
the reservoirs (39 GWh) were not big enough to store this excess energy. Figure
7.5 shows that as soon as the energy reservoir capacity of pumped storage plants
is doubled in Germany, the sensitivity of RES integration to the power rating of
storage plants steeply increases. If Germany is able to store 195 TWh of energy,
then doubling the power rating capacity can help integrate an additional 17 TWh
of var-RES compared to just 1 TWh when its energy capacity is at 39 GWh. How-
ever, expanding the reservoirs of pumped storage plants in Germany is no longer
possible as the country has fully exploited the potential of its mountainous areas.
Therefore other large-scale storage technologies, like CAES and large batteries,
comes very in handy for such situation.
On the other hand, the RES integration in Spain exhibited a high sensitivity to
the increase of the pumped storage plants power rating and no sensitivity to the
energy ratings. This results is very significant as it means that just by installing
more pumps and turbines in the already existing pumped storage plants, the SPS
can absorb up to 52 TWh of intermittent supply, which exceeds the amount of
integrated RES in 2010. Such investments represents low hanging fruits as the
equipment cost 8 of pump storage plants represents less than 5% of the total cost
[66] and the operational benefits of the power system will be significant. The latter
defines the flexibility value, which is presented next.
7.2.2 Flexibility value in Germany and Spain

Similarly to 7.1.3, the flexibility value of each source in each country is calculated
in table 7.6.
Table 7.6: The value of flexibility (€bn per year) in Germany and Spain.
Germany Spain
Parameter NTC SP SE GPS Flex9 NTC SP SE SPS Flex10
Cap. of 2020 20 GW 7.9 GW 39 GWh - 6.6 GW 5.7 GW 1530 GWh -
Double 3.18 0.07 0.15 3.48 1.39 0.82 0.03 2.04
Triple 3.69 0.02 0.19 4.14 1.95 1.23 0.05 2.55
Quadruple 3.32 0.002 0.23 3.97 2.29 1.38 0.07 2.63
Quintuple 2.87 0.01 0.26 3.62 2.53 1.44 0.096 2.74
First to notice is the very low flexibility value of the storage power rating in
Germany and the storage energy rating in Spain which is in line with the insensi-
tiveness of RES integration to those parameters in the respective countries. How-
ever, the GPS flex value exhibit a non-linear pattern with the increase in flexibility
capacity and the NTC flexibility value in Spain is higher than the storage power
8
pump, turbine, generator, transformer
9
GPS Flex = Increase all three capacities simultaneously in the German power system
10
SPS Flex = Increase all three capacities simultaneously in the Spanish power system
rating flexibility value, although the RES integration in the country was sensitive
equally to both parameters. We explain the reasons behind such results in the fol-
lowing:
1. When tripling the flexibility capacities of the GPS, the yearly operational
savings reaches €bn 4.14, however when quadrupling them they decrease
by €bn 0.37 even though the RES integration in Germany has increased
by 3.3%. The reason is that Germany is a very interconnected country and
the more we increase the interconnectedness of the GPS with its neighbors,
the higher the probability that a neighboring country has imported free of
cost German RES, thereby the optimizer is obliged to dispatch more often
conventional plants in Germany. We have not assigned a cost for transmitting
power and since we have assumed that the generation cost of electricity is
the same in all EU countries, the optimizer does not have an incentive to
integrate the RES locally and looks to minimize the operation cost of the
interconnected European system. The question becomes on the willingness
of Germany to invest in a expansion in which the country does not benefit
directly but the EU does. The topic falls outside of our scope of study.
2. Similarly, since we have not defined a transmission cost, the NTC flexibility
value in Spain turned out to be higher than the pump storage power rating
flexibility value, even though the RES integration obtained when increas-
ing each flexibility source separately is the same. The reason is that storing
1 MWh of var-RES will cost 8 €/MWh whereas exporting it to Portugal
or France costs nothing. Nonetheless, the storage power rating flexibility
source has a great value in Spain, from an investment perspective. Consider-
ing only the equipment cost for such plants (0.12 €bn/GW), quintupling this
sources will need an investment of €bn 3.42 which breaks even in less than
2.5 years if all the operational benefits of integrating an additional 52 TWh
are given to pump storage plants.
7.3 CO2 emissions analysis

As we have mentioned in the listed third point of section 6.1.2, the CO2 emissions
from the EPS decreased by 21% compared to the 2010 level, with an RES avail-
ability of 2050 given priority infeed. Even if the capacity of the flexibility sources
of the 2020 EPS are increased by five times to boost the RES integration to their
saturated value (97.5 %), we found that the CO2 emissions would decrease by a
maximum of 47% compared to 2010. This meant that the fleet of the 2020 EPS
had to change to effectively decarbonize the power system. We found that it would
be possible to reach emission reductions of 84% in 2050 without improving the
flexibility of the 2020 EPS but by just phasing out coal and oil power plants and re-
placing them by gas, biomass and geothermal plants. We doubled the gas capacity
and quadrupled the capacity of biomass and geothermal plants to compensate for
the phase out of coal and oil power plants. In the case of Poland and Czech Repub-
lic, the gas capacity had to scaled up by a factor of 8 and 6 respectively to avoid
load shedding, as these two countries were very much dependent on coal power
plants. Figure 7.6 illustrate a graphical representation of the scenarios we have ran
with different power system configurations to find out how the CO2 emissions from
the European power sector will evolve with more renewables in the mix, a change
of the generation mix and an increase of flexibility of the system.
Figure 7.6: CO2 emissions of the European power sector under different power
system configurations.
The Green EPS term that appears on the legend of the figure consist of a Euro-
pean power system with coal and oil power plants phased out and replaced by a mix
of gas, biomass and geothermal. On the right-side of the figure appears the CO2
emissions reduction of each scenario compared to the base year emissions of our
2010 simulation (EPS 2010 RES availability 2010). One notices that the more the
flexiblity value of the Green EPS is increased, the higher the CO2 emission reduc-
tion potential, reaching a maximum of 94% if the flexibility sources are quintupled.
Having previously proved that there is no economic sense in over-tripling the flex-
ibility sources, the CO2 emissions of a coal and oil-free EPS, with three times the
capacity of flexibility sources in 2020, can reach reductions of 91% compared to
the 2010 level.
Part III
Conclusions and
Recommendations
112
114
8.1 The impacts of a high variable renewable supply

• With a var-RES availability reaching 53 % of the electricity demand in Europe in
2050 (5050 TWh) and assuming that O&M cost of variable renewable plants will
be subsidized, the EPS of 2020 can integrate a maximum of 93% of var-RES under
the priority infeed policy. Under an economic dispatch, where curtailing renewable
output is allowed, the RES integration is at 84%. If the O&M cost of VRPs is no
longer subsidized, the integration is reduced to only 74%.
• The operation cost of the same power system configuration under the priority in-
feed is €bn 8.9 more than under a economic dispatch, even though RES integration
was higher. The reason is that economic curtailment allows to ramp up low-flexible
power plants prior to a predicted renewable shortfall, whereas in a priority infeed,
highly flexible plants, which are more expensive, will need to be dispatched to
cover the more rapid shortage. Please note such results are obtained only in a
high-share RES scenario. In a low-share RES scenario (12% of the electricity de-
mand in Europe in 2050), prioritizing the infeed of RES was more economical than
allowing curtailments as the system saves on operation cost €mn 270.
• Moving from a low-RES to high-RES share scenario, the yearly operational sav-
ings of the European power system (EPS) have been evaluated to reach €bn 57
to 95 depending weather wind and solar power plants benefits from a subsidy that
covers their O&M cost or not. If not, the operational benefits are the lower value.
• Power flow exchange patterns will definitely change with more renewable energies
in the generation mix. The traditional labels given to exporting or importing coun-
tries will alter depending on the renewable energy mix installed in the country. For
example, Italy, a well known importer from France and Switzerland, will export
power in the summer to both countries due to the high capacity of PV plants.
• Using the same power system configuration (EPS 2020), the power lines connect-
ing France with its neighbors will not be loaded more heavily under higher-shares
of RES. They are more often loaded close to their NTC capacity under low-shares
of RES.
• The variability of the spot price of electricity will increase with more RES in the
mix but the mean price will decease as well as the electricity generation cost. In-
creasing the share of RES by 7 %, 19 % and 41 % compared to the 2020 RES avail-
ability, the market price of electricity dropped by 9%, 19%, 27% and the electricity
generation cost in Europe dropped by 53%, 72%, 82%, even when considering the
O&M cost of VRPs.
115
8.2 The value of flexibility

• The flexibility of a power system plays a major role in maximizing the integration
of high-shares of RES and thereby minimizing the operation cost of the system.
From a European system-level perspective, the power lines net transfer capacities
showed the highest flexibility value, equaling €bn 9.8 per year if their capacity
were doubled from a total of 170 GW to 340 GW. The value represent the oper-
ational benefits such flexibility expansion results in. The doubling of the pump
storage plants power rating capacity reached a flexibility value of €bn 1.8 per year
and the energy rating reached a value of €bn 0.43. If the three flexibility sources
are simultaneously expanded, the European power system saves yearly €bn 12.04
on operation cost, as 6% more of var-RES are integrated to the grid compared to
the standard system.
• There exist a ceiling to the integration of high-shares of RES. Integrating 100 % of

2675 TWh11 of available var-RES is technically impossible even if the flexibility
value of the EPS is highly increased. We found the ceiling at 97.5 % of RES
integrated if the three flexibility sources were quintupled. However, we found that
if the capacity of the flexibility sources of the 2020 EPS is further than tripled,
the rate of increase of RES integration is no longer significant. When tripled, the
RES integrated to the system increase from a share of 84 % to 94 %, whereas if
quadrupled it reaches 96%, and quintupled 97.5 %. This translated in a decrease in
the incremented flexibility value which indicates that the capacities of the flexibility
source in the power system should not be more than tripled.
• From a country-level perspective, the value of flexibility is unique to the power

system configuration in question and the patterns can differ from the European
system-level flexibility value. The RES integration in Germany was found to be
most sensitive to the expansion of the power lines NTCs and showed a higher
sensitivity to the expansion of its storage energy capacity than its power rating.
Spain was found to be as sensitive to expanding its NTCs than expanding its storage
power ratings capacity, while being totally insensitive to the expansion of its energy
reservoir capacities. Important to note is the critical synergy between doubling
the storage power rating and increasing its storage energy capacity that was only
observed in Germany. If the storage energy capacity is kept the same, doubling
the power rating increases the amount of additional RES integrated in the German
power system by 1 TWh, whereas it the former is doubled as well, then the amount
of RES integrated is tenfold. Each country flexibility value is shown in table 7.6.
Please note that the values illustrated correspond only to the increase in capacity
of each single source alone or their simultaneous increase.
11
53 % of the electricity demand in Europe in 2050
116
8.3 Policy recommendations

• Europe should not hesitate in switching to high-RES shares. Such move should not
be considered only as a sustainable one but also as an economical one. The yearly
operational cost savings on European scale have been quantified to reach at least
€bn 57 under a high-RES scenario compared to a low-RES one.
• A shift in policies is needed to allow for a true system optimization. The priority
infeed should no longer be given to variable renewable plants in a high-share re-
newable scenario. TSOs should have the right to curtail VRPs output for economic
reasons.
• Today’s power flow patterns should not be taken as reference to expand the trans-
mission capacity of the power lines. With more RES in the mix, power exchanges
patterns will definitely change. The expansion should be studied based on different
RES availability supply scenarios and quantifying a highest probabilistic conges-
tion happening frequently between two countries.
• The total power lines net transfer capacities in Europe should not exceed 510 GW,
the pump storage plants power rating should not exceed 141 GW and the storage
energy rating should not exceed 7980 GWh. Above such increase in flexibility ca-
pacities, the additional RES integration benefits are no longer important and there-
fore any savings on operation cost as well. It would be a wasteful over-investment.
• To maximize the RES integration in Europe and thereby minimize the operation
cost of the system, some countries requires investments in different flexibility
sources. For example, Germany should definitely invest in technologies like CAES
or large-scale batteries that allows it to increase its storage energy capacity as there
exist no longer any potential for pump storage plants. Spain, on the other hand,
should absolutely increase the power rating of its pump storage plants as it is not
fully benefiting from its huge energy capacity.
• The European power system can reach emission reduction above 90% only if coal
and oil power plants are phased out and replaced by a mix of gas and biomass
plants. If not, with the highest flexibility value, the emission are reduced by a
maximum of 44% under a high-share var-RES supply.
117
8.4 Future work

We definitely would recommend:
• Finding the threshold of RES supply where the priority infeed is no longer eco-
nomically optimal when operating the European power system
• Conducting an investment analysis that include the capital cost to install the re-
quired renewable plants and study the break-even point of the standard 2020 Euro-
pean power system based on the yearly operational benefits obtained when moving
from a low to a high-share ERS.
• Conducting an investment analysis for the increase of the flexibility parameters in

each EU country and study the break-even point according to the flexibility values
we have quantified.
• Investigating the influence the rise of fuel cost and CO2 emissions pricing has on
the break even.
• Interpolating the hourly externally driven time series into a quarterly basis one and
constraining the pool of power plants to the actual load ramp given in literature.
Define a accurate quadratic cost for ramping power plants and assess the sensitivity
of RES integration to an increase ramping capability of conventional power plants.
• Evaluating the flexibility potential of thermal loads and electric vehicles on a Eu-
ropean system-level scale.
Appendix A
MPC power system dispatch
A.1 Economic dispatch where renewable plants compete

with conventional energies
We have changed in this case the MPC controller so that it has an economic incen-
tive to curtail first the var-RES with highest O&M cost, if the storage reservoirs
are full and if the power can not be exported. For the same German power sys-
tem under the same var-RES infeed in the same dispatch period, figure A.1 differs
from figure 4.3 as there is much less PV power curtailed (orange curve in the sec-
ond sub-plot) and much more wind offshore (blue curve) is curtailed than wind
onshore (green curve). This proves that the MPC effectively minimized the system
cost as it curtailed as last resort the RES with lowest O&M cost.
118
APPENDIX A. MPC POWER SYSTEM DISPATCH 119
Figure A.1: Economic optimum dispatch of the German power system in 2020
under a high-share renewable scenario in which VRE plants compete with conven-
tional technologies.
A.2 Influence of quadratic ramping cost on the power sys-

tem dispatch
We have explained in 7.1.1.3 that the quadratic cost assigned to ramping conven-
tional power plants is crucial in assessing the flexibility measure of their ramping
rates. If we assign a quadratic cost, then we are forbidding any abrasive opera-
tion of the plant and increasing the ramping capability of the plant will not matter.
Figure A.2 illustrate the same dispatch of the same German power system under
the same var-RES infeed in the same dispatch period than in figure 4.3, but in the
former the TSO kept 75% of the fleet turned on and in the latter he kept 25%. That
meant that coal power plant were able to ramp up in one hour 35.3 % of their nom-
inal capacity and once 14.6 % of it, according to table 2.3. Figure A.2 proves that
the dispatch was very much the same as in figure 4.3. However when we removed
the quadratic cost constraint, one notices in figure A.3 a much different dispatch.
We have highlighted in a red circle an important difference we noticed between
both dispatch. Between day 7 to 11, there was no curtailment of renewable energy
if the quadratic cost for ramping power plants was not assigned. This is explained
by the fact that the optimizer turned them off during this period as no red curve can
be observed under the red circle drawn in figure A.3.
Figure A.2: German TSO keeps 75% of the fleet turned on and the quadratic cost
constraint is not changed.
Figure A.3: German TSO keeps 75% of the fleet turned on and the quadratic cost
constraint is removed.
Appendix B
Impact of a high-share renewable

energy supply
A comparison between the NTC line loading in Northern, Central and Easter Eu-
rope under a low-RES energy supply1 and a high-RES energy supply2 is shown.
The message from the French case study in section 6.2 applied also here:
• The expansion of the NTCs connecting European countries should not be

based on the exchange patterns of today as in a high-RES scenario the pattern
of the flows will likely change.
• The common belief that the power system will more heavily loaded with
more RES in the mix is not verified. On average, the European lines are
less loaded close to their NTC in a high-RES than in a low-RES. The figures
below justifies such argument as the density of red-squares is always lower
in high-RES than in a low-RES.
1
12% of the electricity demand in EU in 2050
2
53% of the electricity demand in EU in 2050
123
APPENDIX B. IMPACT OF A HIGH-SHARE RES 124
B.1 Power exchanges in Northern Europe
Figure B.1: Northern European line loading in a low-share renewable supply.
Figure B.2: Northern European line loading in a high-share renewable supply.

B.2 Power exchanges in Central Europe
Figure B.3: Central European line loading in a low-share renewable supply.
Figure B.4: Central European line loading in a high-share renewable supply.

B.3 Power exchanges in Eastern Europe
Figure B.5: Eastern European line loading in a low-share renewable supply.
Figure B.6: Eastern European line loading in a high-share renewable supply.

B.4 Averaged spot market price in each European coun-

try
Figure B.7: Forecast of the spot market price in Europe with increasing RES fixing
the power system configuration to 2020 and the load demand to 2050.
B.5 Averaged generation cost of electricity in each Euro-

pean country
Figure B.8: Forecast of the electricity generation cost in Europe with increasing
RES fixing the power system configuration to 2020 and the load demand to 2050.
Appendix C
Sensitivity analysis results
C.1 Technology specific RES integration results in Spain

We have explained in section 7.2.1 that the overall RES sensitivity in Spain was as
sensitive to the expansion of the power line NTCs than to the storage power rating
of pump storage plants. In fact, since each type of RES has a different variable
infeed, its integration varies slightly differently to the expansion of these two flex-
ibility sources. Figure C.1 clearly illustrate that the solar power plants in Spain
were more sensitive to a storage power rating expansion than the NTCs whereas
the wind power plants energy integration was more sensitive to the expansion of
the latter than the former as illustrates figure C.2.
129
APPENDIX C. SENSITIVITY ANALYSIS RESULTS 130
Figure C.1: PV and CSP energy integration sensitivity relative to power lines NTC
and storage power expansions in Spain.
Figure C.2: Wind onshore and offshore energy integration sensitivity relative to
power lines NTC and storage power expansions in Spain.
Appendix D
Data reference
D.1 Capital investment cost for power lines planning in

22 | TEUSCH, BEHRENS & EGENHOFER
Europe [1]
Table 4. Selected, planned interconnector projects between Nordic and other northern European countries
Countries Name Capacity Year Cost est. Owner/operator Additional information
From To (MW) (est.) (€ mn)
DK DE Reinforcements I 500 2012 NA Energinet.dk and TenneT
DK DE Reinforcements II 500 2017 NA Energinet.dk and TenneT
DK DE Krieger's Flak 900 2018–20 1,200 50Hertz Transmission (Elia), 600 MW HVDC, 300 MW AC; capacity only
Energinet.dk available for trade if not used by wind farms;
Sweden might join at a later stage; own cost
estimate based on a feasibility study
DK NL COBRAcable 700 2016 456 TenneT, Energinet.dk Subsea HVDC
FI EE Estlink2 650 2014 320 Fingrid and Elering Subsea HVDC
NO DE Nord.Link 1,000 2018–21 1,544 Statnett and TenneT Subsea HVDC; regulated investment
NO DE NorGer 1,400 2016? 1,400 Statnett Subsea HVDC cable; the Commission did not
grant an exemption as a merchant line; the future
is uncertain
NO NL NorNed2 700 >2021 683 Statnett and TenneT Subsea HVDC, postponed as Statnett prioritises
projects with Germany and the UK
NO UK NSN 1,000 2018–21 1,740 National Grid and Statnett Subsea HVDC, possibly a wind farm connection
mid-way (Dogger Bank)
NO UK NorthConnect 1,400 <2020 1,740 Agder Energi, E-CO, Lyse, SSE Subsea HVDC; capacity might be extended to up
(Scottish and Southern Energy), to 2,000 MW; merchant; own cost estimate based
Vattenfall AB on an NSN project
SE LT NordBalt 700 2015–16 553 Svenska kraftnät, LITGRID Subsea HVDC
Turtas AB
Total 9,450
Note: For investment cost estimates in currencies other than €, the following exchange rates were applied: DKK 1 = €0.13, NOK 1 = €0.13, SEK 1 = €0.11; in cases of contradictory
estimates, the most recent one is reported.
Sources: ENTSO-E (2010, 2011a), P&T (2010), the reports of national regulators and press releases of project developers.
Figure D.1: Selected interconnector projects in northern Europe [1].
131
46 | TEUSCH, BEHRENS & EGENHOFER
APPENDIX
Appendix 6. Selected D. DATA
interconnector REFERENCE
projects in northern Europe 132
Table A7. Selected interconnector projects in northern Europe
Countries Name Capacity Year Cost Owner/operator Additional information
From To (MW) (est.) (€ mn)
GB NL BritNed 1,000 2011 600 National Grid and TenneT 260 km submarine cable, HVDC; merchant;
operating since April 2011, with more cables planned
by 2020
SE FI Fenno Skan 2 800 2011 300 Fingrid and Svenska HVDC subsea
Kraftnät,
DK NO SK4 600 2014 443 Statnett and Energinet.dk 450/500 kV DC, regulated
PL LT LitPol 1,000 2015-20 237 PSE Operator S.A. and HVDC, overhead, to be completed in two steps
Lietuvos Energija AB
SE NO Southwest Link 1,200 2016-17 - Statnett and Svenska Combination of HVAC and HVDC
Kraftnät
BE GB NEMO 1,000 2016-18 - National Grid International Planned as cap and floor
Limited (NGIL) and Elia
Sources: P&T (2010) (based on Nordel).
Figure D.2: Selected interconnector projects in northern Europe [1].
D.2 ENTSO-E net transfer capacities matrix

The net transfer capacities of each line connection in Europe as calculated by the
ENTSO-E in [61] are illustrated in the next page.
Prepared by RGCE SG Network Models and Forecast Tools
Indicative values for Net Transfer Capacities (NTC) in Europe

Summer 2010, working day, peak hours (non-binding values)
W A R N I N G: Legend :
1200
Version These indicative values have been computed by extrapolation from NTC in MW - Value agreed by both countries
ENTSO-E country
FINAL standard situations, in order to evaluate the transfer capacity 600 NTC in MW - Different values are estimated between the two countries involved. The lower value is shown on top and the
through a single interface for a typical exchange situation 800 CZ country providing the higher value is specified,
6 July 2010 (European Reference Case Map also available on ENTSO-E Other country
Website). Thus these figures are only indicative and they are not 600 NTC in MW - Value provided by only one country
cumulative (they cannot be summed up). Maximum export and SE The country providing no value is specified.
import values per country also intend to clarify this.
NRL no realistic limit (5)
From Maximum
IE/NI GB MA PT ES FR BE NL LU DE DKw DKe NO SE FI CH IT AT SI PL CZ SK HU GR RO HR BA RS ME MK AL BG UA LV LT EE BY RU
(9) (8) (1) (8) (2) (11) (13) Import
To (6) in MW
IE/NI 410
GB 80 2000
MA 900
MA
PT 1200
ES 600 1200 1200

MA
FR 2000 500 1300 3200 1100 870
BE 2900 2200
NL 2300 4000 700 3850 (4)

NL
LU 980
DE (9) (8) 2600 3900 NRL 1500 (4) 550 600 4400 1600 1200 (4) 2100 (4)
PL (16)
DKw (1) (8) 950 (4) 950 340

1000 NO 680 SE
DKe (2) 550 1300
NO 700 950 3700

NO 1000 NO 3795 SE
SE 600 370 1700 3545 1650 0

740 SE 3600 NO FI 600 SE
FI 2050 350
FI FI
APPENDIX D. DATA REFERENCE
CH 3000 2060 1440 540

1000 AT
IT (11) 2400 3460 200 330 500
AT 1600 1000 70 900 800 350

2110 CZ 600 AT
SI 120 900 700

1000 SI
PL (13) 800 (4) 600 800 500 0 (14)

PL (16) PL(16) PL (16)
CZ 800 (4) 600 1900 1100

900 AT PL (16) 1350 CZ
SK 600 2100 500 400

PL (16) 2500 CZ 600 SK
HU 500 1150 500 500 600 800

700 AT 1250 HU 700 RO 600 HR 1150 HU
GR 500 250 100 800

200 GR
RO 600 300 400 400 1600
HR 800 1000 450 350

1000 HR 600 HR
BA 550 350 450

600 BA
RS 600 600 400 500 400 200 210 400

500 RS
ME 400 400 220
MK 100 450 450
AL 100 210 200
BG 100 400 100 50
UA 400 650 200

1000 HU
LV 1250 500 350 900

550 LV
LT 1100
Figure D.3: Indicative ENTSO-E net transfer capacities in Summer 2010.

EE 350 500 500
FI 550 EE
BY
133
350 500
RU
Maximum 1100
3850 8500 (3)
Export (4) (4)
1700 600
(14)
in MW
Footnotes : (1) - Denmark West (8) - NTC values should be considered separately, are not cumulative and simultaneous
(2) - Denmark East (9) - Structure of GCB : Amprion, EnBW TNG, TPS, 50Hz Transmission, Creos (LU), TIWAG-Netz (AT), VKW-Netz (AT)
(3) - Maximal cumulated value from DE to NL, FR and CH (10) - Regulated capacities
(4) - Depending on wind situation in Germany (11) - Provided by Austria-France-Italy-Slovenia-Switzerland
(5) - When the evaluation of a transfer limit requires to use a set of assumptions which are too far from (13) - Because of meshed system in the region, PL provides only one interdependent value (for each direction) for the whole polish profile. The value meet requirements of coordinated
usual or foreseeable situations (leading to high inaccuracies), the NTC value has been replaced by auctions in the region.
"no realistic limit". (14) - This refers to the maximal value from/to DE, CZ and SK
(6) - Countries are listed in the matrix from the West to the East of Europe (15) - This refers to the maximal value from/to FR, DE, AT
_ (16) - PL provides only the interdependent value for the whole polish profile. The value is given in the Maximum Export\Import in MW column\row
2.11.2009
Indicative values for Net Transfer Capacities (NTC) in Europe

Winter 2009-2010, working day, peak hours (non-binding values)
W A R N I N G: Legend :
Version 1200
NTC in MW - Value agreed by both countries
These indicative values have been computed by extrapolation from ENTSO-E country
2nd November standard situations, in order to evaluate the transfer capacity through 600 NTC in MW - Different values are estimated between the two countries involved. The lower value is shown on top and the
a single interface for a typical exchange situation (European 800 CZ country providing the higher value is specified,
2009 SEE TSO country
Reference Case Map also available on ENTSO-E Website). Thus these 600 NTC in MW - Value provided by only one country
figures are only indicative and they are not cumulative (they cannot SE The country providing no value is specified.
Other country
be summed up). Maximum export and import values per country also
intend to clarify this. NRL no realistic limit (5)
From (6) Maximum

IE GB MA PT ES FR BE NL LU DE DKw DKe NO SE FI CH IT AT SI PL CZ SK HU GR RO HR BA RS ME MK AL BG UA LV LT EE BY RU
(9) (8) (1) (8) (2) (11) (13) Import
To (6)
in MW
IE
GB
MA 900
MA
PT 1500
ES 600 1300 1300

MA
FR 500 2300 3050 2300 995
BE 3400 2400
2400 3850 3850 (4)
NL
LU 980
2800 3000 NRL 1500(4) 550 610 3200 2000 1100 DE(4) 2300 (4)
DE (9) (8) PL (16)
DKw (1) (8)

950 (4) 950 680
NO SE
DKe (2) 550 1300

SE
NO 950
NO
SE 600 740 1700 0

SE SE
FI
APPENDIX D. DATA REFERENCE
CH 3200 1500 1810 470

1200 AT
IT (11) 2650 4240 220 430 500

650 SI
AT 2200 1200 85 900 1000 300

285 AT 2180 CZ 800 HU
SI 160 900 1000

650 SI
1200 DE(4) 600 800 500 0 (14)
PL (13) PL (16) SE PL (16) PL (16)
CZ 800 (4) 600 2000 1000

1200 AT PL (16)
SK 600 1700 600 400

PL (16)
HU 100 1250 1100 800 600 650

400 AT 1400 HU 1500 HU 800 UA
GR 500 350 150 600
RO 600 300 600 400 1900

550 RO
HR 1000 1000 700 350
BA 570 350 380

400 BA
RS 600 600 400 400 400 350 200 350

480 BA AL 500 BG
ME 500 400 250
MK 300 350 450
AL 200 200 250

AL
BG 300 600 300 200

250 MK
UA 400 650 300

800 HU 400 UA
LV
LT
EE
BY
Figure D.4: Indicative ENTSO-E net transfer capacities in Winter 2009-2010.

RU
134
Maximum
3850 1000
Export 8500 (3) (4) 1900
(4)
(14)
in MW
(1) - Denmark West (7) - This refers to the maximal value from GR to BG, MK and AL
Footnotes : (2) - Denmark East (8) - NTC values should be considered separately, are not cumulative and simultaneous
(3) - Maximal cumulated value from DE to NL, FR and CH (9) - Structure of GCB : Amprion, EnBW TNG, TPS, Vattenfall Europe Transmission, Creos (LU), TIWAG-Netz (AT), VKW-Netz (AT)
(4) - Depending on wind situation in Germany (10) - Regulated capacities
(5) - When the evaluation of a transfer limit requires to use a set of assumptions which are too far from (11) - Provided by Austria-France-Italy-Slovenia-Switzerland
usual or foreseeable situations (leading to high inaccuracies), the NTC value has been replaced by (13) - Because of meshed system in the region, PL provides only one interdependent value (for each direction) for the hole polish profile.
"no realistic limit". The value meet requirements of coordinated auctions in the region.
(6) - Countries are listed in the matrix from the West to the East of Europe (14) - This refers to the maximal value from/to DE, CZ and SK
_ (15) - This refers to the maximal value from/to FR, DE, AT
(16) - PL provides only the interdependent value for the hole polish profile. The value is given in the Maximum Export\Import in MW column\row
Version 1.5
Bibliography
[1] Christian Egenhofer, Jonas Teusch, and Arno Behrens. The benefits of in-
vesting in electricity transmission - lessons from northern europe. Center of
European Policy Studies, ISBN 978-94-6138-161-3:27, 51, January 2012.
[2] EuroObserv’ER. Photovoltaic barometer. le journal du photovoltaique, Avril,

2012.
[3] EuroObserv’ER. Wind energy barometer. le journal de l’eolien, February,

2012.
[4] ABB Adrian Timbus. Managing wind power plants a view on control possi-
bilities. In Swiss electromagnetics research and engineering centre, Decem-
ber 2012.
[5] Andreas Ulbig, Stephan Koch, and Göran Andersson. The power nodes mod-
elling framework - modeling and asessing the operational flexibility of hydro
power units. In XII Symposium of specialists in electric operational and ex-
pansion planning, 20-23 May 2012, Rio De Janeiro, Brazil.
[6] Hugo Chandler. Harnessing Variable Renewables. IEA, 2011.
[7] Guiseppe Lorubio. Flexibile generation: Backing up renewables. Technical

report, EurElectric, 2011.
[8] Pierre Schlosser and Arta Denina. Hydro in europe: Powering renewables.
Technical report, EurElectric, 2011.
[9] Prof. Dr. Uwe Leprich. Preissenkende effekte der solarsstromerzeugung auf
den borsenstrompreis. In Kurzstudie im Auftrag des Bundesvrebandes Solar-
wirtschaft (BSW). Institut fur ZukunftsEnergieSysteme, Januar 2012.
[10] European Wind Energy Association. Wind in power: 2012 european statis-
tics, February 2013.
[11] Kai Heussen, Stephan Koch, Andreas Ulbig, and Göran Andersson. Unified
system-level modeling of intermittent renewable energy sources and energy
storage for power system operation. IEEE SYSTEMS, 2011.
135
BIBLIOGRAPHY 136
[12] Sytze Dijkstra, Enrique Gaxiola, Frans Nieuwenhout, George Orfanos, Ozge
Ozdemir, and Adreaan van der Welle. European scenario synthesis to be used
for electricitz transmission network planning. IEEE, 2011.
[13] TradeWind. Integrating wind developing europe’s power market for the large-
scale integration of wind power. pages 23–29, February 2009.
[14] IEA. Electricity information 2012. IEA statitstics, ISBN 978-92-64-08418-

6:102–108, 303, 445, 463, 697, 787, 805, 2012.
[15] ENTSO-E. Detailed monthly production in france in 2010, June 2013.
[16] Wikipedia Encyclopedie. List of pumped-storage hydroelectric power sta-

tions.
[17] Swiss Federal Office of Energy.
[18] Internation Energy Agency and Nuclear Energy Agency. Projected costs of
generating electricity. ISBN 978-92-64-08430-8:61–65, 2010.
[19] IEA. Co2 emissions from fuel combustion. IEA, ISBN 978-92-64-10283-
5:51, 2011.
[20] IRENE 40. Internal report. task 2.3 - scenario synthesis, 2012.
[21] Christina Figueres. Environmental issues: Time to abandon blame-games and

become proactive. The Economic Times, 2012.
[22] State of the Environment Report. European Environment Agency, 2010.
[23] European Comission Climate Action. Roadmap for moving to a low-carbon

economy in 2050, March 2011.
[24] Göran Anderson. Dynamics and Control of Electric Power Systems. ETH
Zürich - Power Systems Laboratory, February 2012.
[25] Bp statistical world energy review 2011, August 2011.
[26] Xu Hongua, Charlie Dou, Wang Sicheng, and Lv Fang. National survey re-
port of pv power applications in china. China National Photovolotaics Status
Report, 2012.
[27] EurElectric. Key statistics. 2012.
[28] Ian Steadman. Germany sets solar record, meets half of electricity demand,
May 2012.
[29] Prof. Dr. Bruno Burger. Electricity production form solar and wind in ger-
many in 2012. Fraunhofer Institute for Solar Energy Systems.
BIBLIOGRAPHY 137
[30] Anders Lorenzen. Breaking: Denmark records highest wind power output.
[31] Wikipedia encyclopedia. Electric generator.
[32] Zachary Shahan. China & byd launch largest battery energy storage station
in the world. Clean Technica, January 3, 2012.
[33] Louis Rondeau. Profitability assessment of battery energy storage systems.

Master’s thesis, ETH Zürich, August 2012.
[34] David Dallinger, Gerda Schubert, and Martin Wietschel. Integration of inter-
mittent renewalbe power supply using grid-connected vehicles - a 2030 case
study for california and germany. Working Paper Sustianability and Innova-
tion, April 2012.
[35] IRENE-40. Methodology for determining impact and value of dsp on system
infrastructure development.
[36] GOV.UK Department of Transport. Vehicle licensing statistics q2 2012, 13

September 2012.
[37] B. Roberts. Capturing grid power. IEEE Power and Energy Magazine, pages
32–41, july-august 2009.
[38] Fernando De Samaniego Steta. Modeling of an advanced adiabatic com-

pressed air energy storage unit and an optimal model-based operation strategy
for its integration into power markets. Master’s thesis, ETH Zürich, 2010.
[39] Josef Auer. State-of-the-art electricity storage systems. In Current Issues

Natural resources. Deustche Bank Research, 2012.
[40] John Wiley & Sons. Power Systems in Emergencies - From Contingency
PLanning to Crisis Management. Knight U.G, 2001.
[41] Ride Energy Storage & Grid Services. Caes technology: History.
[42] EurElectric: John Scowcroft and Susanne Nies. Eurelectric views on demand-
side participation. August 2011.
[43] Rainer Bacher. Optimization of liberalized electrical energy systems. Depart-

ment of Electrical Engineering and Information Technologies, 2011.
[44] Moma Modellstaft Mannheim. E-energy-project smart city mannheim, 2011.
[45] European Transmission System Operators. Procedures for cross-border trans-

mission capacity assessments. October 2001.
[46] Joshua S Hill. Denmark renewable energy generation. Clean Technica,

September 28, 2012.
BIBLIOGRAPHY 138
[47] EnergieNet.DK. Wind power to combat climate change, how to integrate

wind energy into the power system? Green thinking in Denmark, 2010.
[48] DG Energy Dinko Raytchev. Quaterly report on european electricity markets.

European Commission, 5, Issue 1:5, January 2012- March 2012.
[49] SMA Power Control Module. Multi-function interface for pv inverters. Tech-
nical report, July 2012.
[50] Morten Grud Rasmussen, Gorm Bruun Andresen, and Martin Greiner. Stor-
age and balancing synergies in a fully or highly renewable pan-european
power system. Energy Policy, September 2012.
[51] Wietze Lise, Jeroen van der Laan, Koen Rademaekers, Frans Nieuwenhout,
and Christian Kirchsteiger. Assessment of the required share for a stable eu
electricity supply until 2050. European Comission, October 2011.
[52] Gary Keane. Options for low-carbon power sector flexibility to 2050. Tech-
nical report, Poyry, October 2010.
[53] Roman Kapuscinski Owen Q.Wu. Curtailing intermittent generation in elec-

trical systems, June 2012.
[54] Philip A.Jonas. Predictive power dispatch for 100using power nodes model-
ing framework. Master’s thesis, ETH Zürich, July 2011.
[55] Andreas Ulbig and Göran Andersson. On operational flexibility in power

systems. In IEEE PES General Meeting, San Diego, USA, 2012.
[56] Arthouros Zervos, Christine Lins, Lucie Tesniere, and Eleanor Smith. Map-
ping renewable energy pathways towards 2020. European Renewable Energy
Council, page 6, March 2011.
[57] GE Enegy Stefan Hartge. Integration wind energy into the grid, February
2009.
[58] EurElectric. Power statistics & trends 2011 - full report. pages 58–75, 2011.
[59] ’Data provided by ENTSO-E’. Detailed monthly production in germany for

the year 2010.
[60] Stephen Lacey. Groundbreaking solar plant in spain generates 24 hours of

power. Climate Progress, June 2011.
[61] ENTSO-E European Network of Transmission System Operators. Ten-year

network development plan 2010-2020. June 2010.
[62] Bundesamt fur Energie BFE. Statistik der wasserkraftanalgen der schweiz.
2010.
BIBLIOGRAPHY 139
[63] IEA. End-use petroleum product prices and average crude oil import costs.
September 2012.
[64] VGB Powertech. Investment and operation cost fitures - generation portfolio
survey. November 2011, page 6.
[65] Herman Migliore Tim Anderson. SYSC - 505 Power System Economic Dis-
patch. Portland State University, 2010.
[66] Barry Brook. Pumped-hydro energy storage - cost estimates for a feasible
system. Brave new climate, April 2010.

Comaty MA 2013

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Comaty MA 2013

Uploaded by

Copyright:

Available Formats

eeh power systems

Modeling and Simulation of the European

EEH – Power Systems Laboratory

Expert: Prof. Dr. Göran Andersson

Zürich, August 31, 2013

List of Tables xiii

2.3 Swiss Federal Office of Energy . . . . . . . . . . . . . . . . . . . 35

3 Modeling power nodes 40

4 The dispatch simulator 49

II Results and discussion 62

5 Simulation vs Real-World dispatch results 64

6 The impacts of a high-share of renewable energy supply on the Euro-

6.1.1 Scope of analysis . . . . . . . . . . . . . . . . . . . . . . 72

7 Sensitivity analysis results 83

III Conclusions and Recommendations 97

A MPC power system dispatch 103

B Impact of a high-share renewable energy supply 108

C Sensitivity analysis results 111

D Data reference 112

TSO Transmission system operator

1.1 Installed PV Power Capacity in the European Union at the end of

4.1 Basic structure of the MATLAB-based simulation tool. . . . . . . 50

5.1 Power node dispatch in Austria vs IEA references in 2010 [14]. . . 65

6.1 Influence of renewable policies on renewable integration results. . 76

A.1 Economic optimum dispatch of the German power system in 2020

B.1 Northern European line loading in a low-share renewable supply. . 108

C.1 PV and CSP energy integration sensitivity relative to power lines

D.1 Selected interconnector projects in northern Europe [1]. . . . . . . 112

1.1 Characteristics of exemplary pump hydro plants in Europe [16] . . 9

2.1 Overview of the five IRENE-40 demand and generation scenarios

3.1 Characteristics of the Power node i = 1, which represents the ag-

5.1 Power node dispatch vs IEA reference values in 2010. . . . . . . . 68

6.1 Comparison of dispatch policies. . . . . . . . . . . . . . . . . . . 73

6.4 NTC installed in the European power system of 2020 . . . . . . . 77

7.1 Operation cost of system update (bn € per year). . . . . . . . . . 85

Germany appears to be the leading EU country with more than 29 GW of Wind

1.2 Sources of operational flexibility in the grid

1.2.1 Ramping capability of generators

1.2.2 Energy storage

Pump hydro storage

Table 1.1: Characteristics of exemplary pump hydro plants in Europe [16]

Location Reservoir (Mm3 ) Head (m) Capacity (GW) Energy (GWd)

A sample calculation for the PHS plant Grand’Maison in France is shown2 .

Compressed air energy storage

1.2.3 Demand side participation

can be either manually done or automatically controlled via innovative controller

1.2.4 Net transfer capacity of cross border flows

1.2.5 Controllability of wind and solar power units

To provide control options according to grid codes requirement, the WF con-

SMA Solar Technology AG also came up with a innovative Power Control

Figure 1.6: Existing operational flexibility in power systems [5].

1.3 Literature survey on the integration of high-shares of

Figure 1.7: Flexibility Assessment Method of the IEA [6].

Height of bar shows deployment potential

will not be fully available (see lower part of figure)

Score: High Medium Low

18 The availability of flexible resources for the balancing task is

Figure 1.9: Flexiblity of conventional power generation technologies [7].

Figure 1.10: Evolution of generation in Spain on March 3 2010 [7].

Figure 1.13: Developed and remaining technically feasible hydropower potential

true potential it can play in balancing variable renewables. Contrary to popular

1. Taking into consideration the flexibility option of curtailing renewable plants

2. Developing a method that unifies traditional power system operation with

4. Modeling the cross-border transmission constraints between EU countries.

Technology(MW) Austria Belgium Bulgaria CzechRepublic