You are on page 1of 5

An intelligent tool aimed to policy makers:

combining costs, risks and sustainability


Leontina Pinto
ENGENHO
Rio de Janeiro, Brazil
leontina@engenho.com
AbstractThis paper proposes a contribution for the optimal
expansion planning concerning the three main conflicting goals:
sustainability, economy and supply risks. The proposed
approach aims to build an intelligent function that acts like a
bridge between technicians and society, translating technical
issues into relationships between desired targets in a
transparent, straightforward platform. The results may support
decision-making processes in planning, operation and, most of
all, public policies.
Index TermsRenewable penetration,
Expansion planning, Risk Analysis

I.

Market

design,

INTRODUCTION

Historically, society swings, almost like a pendulum,


between low-cost and clean energy. A sustainable world
would rely on renewable energy usually eolic, small
hydroelectric or solar plants. However, the minimization of
the carbon footprint comes with a price, not always clear to
the general public and sometimes only discovered in crisis
moments. Moreover, the risks associated to some decisions are
not always (almost never) perceived; for instance, what is the
supply risk associated to the eolic, or photovoltaic penetration
in the energy matrix?
II.

OBJECTIVE

This paper addresses these concerns through the


construction of a framework relating costs/risks/emissions,
easily understood by the society, yet containing all the
characteristics and restrictions of a complete planning and
operation problem. The main idea is to translate technical
complexities into an intelligent, simple, straightforward
function that summarizes the inter-relations between the main
objectives. It will be possible to answer questions such as

What is the cost of a sustainable energy matrix?


What is the associated supply risk?
How would a lower risk level impact on
sustainability?
How to adjust a balanced risk/cost/sustainable energy
matrix?

III.

THE OPTIMAL EXPANSION PLANNING MODEL

The electrical energy optimal expansion planning has been


extensively covered by the technical literature. Ref. [1] offers
an extensive state of the art, and Ref. [2-4] describe in detail
the many aspects of the problem, embracing energy and
economics. Moreover, Ref. [5] extends an eye to the future,
discussing system and markets recent challenges and
evolution.
Modern planning is essentially multi-objective. It is
necessary to account for uncertainty, minimize costs and risks,
maximize sustainability. Again, the literature offers many
different approaches to mitigate risks [6] and search robust
solutions [7].
The optimal expansion model used in this work has been
described in [8,9] and will be therefore only briefly
summarized here. It must be stressed, however, that the main
contribution of this paper does not rely on the optimal
portfolio, but in its translation into a decision-making
framework. The interested reader may find formulation
details in the cited works, among other references.
A. Basic minimum cost formulation
This paper presents the basic optimal portfolio formulation
designed and implemented for Brazil that is, for a hybrid
hydro-eolic-thermal system. Further extensions, including
photovoltaic or biomass sources, are straightforward and can
be easily implemented.
The overall planning problem may be written as
Min c ( x ) E ( d ( y s , ))

s / to

A(x ) b

(1)

H (y s , ) (x )

x are the investment variables to be built on instant


here represented by

hydroelectric plants maximum storage volumes V

and turbined energy Q


thermal generators capacity T
eolic installed capacity W
line flow limits F

ys, are the operation variables on scenario s, instant


eolic generation W s ,
hydroelectric generation Q s,
thermal generation T s ,
line flows F s ,
eventual deficits D s , on load supply

c(x ) are the unit costs associated to expansion variables x


during instant
E(d(ys, )) are the expectations of the costs associated to
operation variables y along future operation scenarios s
and time intervals given by
E (d ( ys , )) ps d ( ys , )

(2)

comprehensive model should also account the total emissions


associated to resource construction (steel, oil, logistics, etc.),
this paper will target the emissions associated to the operation
mostly thermal units. Huge reservoir impacts will not be
modeled, as they are (at least) controversial and are not
expected to be constructed within the planning horizon.
Moreover, new hydroelectric plants emissions are not
expected to produce a significant impact to deforestation:
Brazilian modern reservoirs are now constructed on already
flooded areas, with minor impact on fauna/flora.
Although the minimum-emission expansion could be
modeled as an objective to the optimum planning which
would therefore be characterized as a multi-objective problem
[10] this paper accommodates sustainability targets, as the
maximum emission levels established, for instance, by
international agreements or national policies. Under these
hypotheses, expected emission levels are written as

where is the expected emission levels in instant given


by

A(x ) b are the investment constraints (for instance,


budget limits or environmental restrictions)

t ps bqi Qsi , bwjWs ,j btk Tsk,

where ps is the probability of scenario s

(5)

(6)

H(ys, )
+(x ) are the operation constraints for
scenario s and instant (for instance, system physical
requirements/characteristics or physical limits associated
to expansion variables x).

where bqi, bwj, and btk are the unit emissions associated to
the operation of each hydroelectric plant i, each wind plant j
and each thermal plant k. These figures are usually known
from international references and previous studies; this work
uses the emission table published by the Brazilian Mines and
Energy Ministry, consistent with the International Panel on
Climate Changes [11].

It is important to notice that the operation constraint set


contains, among others, the system load balance

Emission levels in instant may be constrained by a


maximum desired level *

G
g
s,

Ls ,

s , t

(3)

where Gsg, is the total generation of source g in scenario s,


instant . As our system is composed by hydroelectric,
eolic and thermal units, equation (3) corresponds to

Qsi , Ws ,j Tsk,
i

Ls , Ds ,t

(4)

stating that, for each scenario s and each instant the sum
of each hydroelectric plant i, each wind plant j and each
thermal plant k equals the total system load minus eventual
deficits.
Of course, equation (4) may be granularized, according to
system representation and study goals, from systems
regions to feeders or buses. This extension may be easily
written and will not be discussed in this paper.
B. Sustainability
The search of sustainability aims to minimize the total
system emissions. While recognizing that a more

(7)

C. Risk constraints
Acceptable risk levels of unserved load (risk constraints)
correspond to additional constraints to problem (1) defining
the maximum expected deficit expectation at instant
pD

s,

(8)

D. Problem solution
The optimal portfolio (1-8) corresponds to a large-scale
stochastic linear programming problem, solved by a standard
optimizer under a Matlab framework. A careful
implementation, taking advantage of problem structure and
sparsities, allow the overall solution of the entire problem
(some thousand variables, one million constraints) within a
few minutes in a normal I6 computer.
IV.

SENSITIVITY ANALYSIS

The marginal cost of the variations on acceptable emission


levels in instant is given by

e,

(9)

where e, is the dual multiplier associated to emission


limit constraints
Similarly, the marginal cost of variations on acceptable
risk levels in instant is given by

(10)
r,

where , is the dual multiplier associated to the

Figure 3 Cuts overlap

maximum unserved energy constraints


Of course, sensitivities are calculated around the optimum
and may not be valid for the entire space of solutions.
Fig. 1 illustrates the typical results of a unidimensional
function construction by Monte-Carlo sampling, showing the
construction of the cost function around different samples.
Although sometimes not intuitive, the number of linear
segments for a real-life problem is usually small (say, less
than 20 for a realistic study)
Figure 4 Final envelope function

V.

THE OVERALL PROCESS

The overall process may be summarized as


For each instant of the horizon of study:
1- Sample a parameter set (in our case, a pair
Figure 1 Monte-Carlo construction of an unidimensional cost function

Fig. 2 shows a typical sampling result, highlighting the


sensitivity cuts generated by the different samples. The
algorithm may yield (and it usually does) similar (replicated)
cuts along the process, which are eliminated by a screening
feature.

2- Solve problem (1-8) and evaluate sensitivities e,and


,
3- Go back to 1 and continue sampling.
Establishing a convergence criterion is not intuitive. As
our goal is to derive a function and not a variable, statistic
techniques, based for instance on a desired indicator mean and
variance, would not make sense.
This work takes as a convergence criterion a minimum
number of samples without significant improvements on the
function surface. In other words, the process stops when it is
not able to produce any new sensitivities able to modify the
envelope function As a rule of thumb, practice has shown that
a minimum number of samples (say, 500) without producing
new information signals the end of the process.

Figure 2 Sampling result

Finally, it is interesting to observe that different cuts may


overlap each other. For instance, a superficial analysis, shown
in Fig. 3, could lead to the conclusion that point x correspond
to a set of different cuts.
In fact, the Duality Theory [12] states that the optimal
solution corresponds to the maximum sensitivity value in
other words, the desired function is given by the cuts
envelope, as depicted in Fig. 4.

VI.

CASE STUDY

The described model was applied to a realistic case study:


the multi-objective Brazilian expansion planning that is,
balancing the sustainability/risks/costs requirements
targeting year 2023. As discussed, our focus will not be the
optimal solution, but the intelligent function that
encompasses all the necessary information for the public
policy makers, relating costs, risks and sustainability.

A. System Description
We followed the official system representation [13], which
divides the system into regions that share similar resources
and risks. Fig. 5 depicts main load locations (Southeast), eolic
plants (beige) and small hydros (blue): wind and water
harmonically cover the country, as a teamwork. Dark
zones correspond to sparsely populated areas, where most
large plants are located, usually served by the integrated grid.

It is interesting to observe that (due to huge government


tax waivers policies) wind energy costs are actually very low.
In fact, large hydroplants and wind energy are the most
competitive sources, and there is no real conflict between
economy and sustainability). In other words, a cost-effective
expansion will follow the green way.

Figure 7 Average expansion costs by source (as of 2013)

Figure 5 Brazilian hydroeolic geographical complementarity

The initial configuration (march, 2015) is synthesized in


Fig. 6; generation, load and system data were obtained from
the official data [14].

Current decision-making process would take the dreamgreen solution as a basis for the expansion plan; not
surprisingly, Brazilian expansion has been dominated by large
hydroelectric plants and wind farms.
However, it has been increasingly recognized that the
green option may hide some dangers. Fig. 8 shows the
Brazilian renewables dynamic for the last years, normalized
for better comparison. It is possible to observe the occurrence
of climatological critical scenarios such as year 1994
where solar radiation, wind velocity and hydro inflows are
low, and would lead to supply risks. In other words, a pure
minimum-cost expansion may not be the best solution.

Figure 6 Brazilian source portfolio march/2015

Control variables correspond to new plants (hydro, wind


and thermal; photovoltaic penetration is, at this time, still
under incentives and will not be considered in this study.
Constraints correspond to load supply and physical constraints
for energy and peak levels at each month.
B. Uncertainties
The proposed model is general; uncertainties are modeled
independently, by an external tool, and resulting scenarios will
be treated regardless of its origin.
This particular case study uses a set of scenarios obtained
from specialized joint climatological hydro/eolic forecast
models beyond the scope of this paper the interested reader
may found the details in [15]. Load uncertainties and system
failures, though easily accommodated, will not be focused in
this study, for better consistency with official methodologies
which only incorporates climatological uncertainties.
C. Risks, costs and emissions
Fig. 7 displays the incremental
(US$/MWh) for each available source.

expansion

costs

Figure 8 Normalized monthly series of solar radiation (red, right axis),


river inflows (blue, left axis)) and wind speed (green, left axis) 1990-2011,
Northeastern Brazil (pinned on map)

It is necessary, therefore, to choose between some possibilities


1. taking this risk and stick with a green and economic
plan
2. install thermal units as a backup for climatological
variability risks
3. install a surplus of green units (eolic or hydrothermal)
in order to increase offer and account for availability
reduction when and where necessary
This decision encompasses a broad range of aspects, from
sustainability to scarce economic resources management.
Therefore, it should not, in principle, stay within technical
domain. Instead, it is a matter of public policy, to be discussed
with the whole society.

Policy makers and most leaders need a concise, clear tool


able to measure without losing precision the
consequences of each possible alternative.
To reach this goal, we applied the overall Monte-Carlo
algorithm described in Section V, sampling possible
hydroeolic availabilities and sustainability/risk levels. Fig. 9
shows the resulting inter-relation function between risks
(measured as expected curtailed load), emissions (measured in
expected millions of equivalent tons of Co2 and costs
(measured in additional US$ millions to achieve desired
targets).

VII. CONCLUSIONS
More than simply calculating the best expansion planning,
this paper aims to offer policy makers and society leaders an
efficient, precise and accurate tool to synthetize a multiobjective problem and express system complexities into a
clear tool that summarizes the advantages and limitations
(the consequences, in short) of each policy, and adjust
alternative solutions to desired requirements and goals.
This intelligent function may be the input to further
decision processes, carried out by non-technical policy makers
using specific management models, without lying on
incomplete perceptions, but on concrete, reliable information.
We hope to contribute to a rich, concrete discussion, based
on a wider comprehension and understanding.
REFERENCES
[1]

[2]

[3]
Figure 9 Expansion cost reduction as a function of emissions/risks

The derived function may be seen as a simple yet


complete tool for a conscious, solid decision that will not
lead to unsuspected risks or unforeseen problems. It is
possible to know, precisely, the increase of risks and/or costs
associated to any degree of emission levels. Moreover, it will
be possible to establish a complete plan, choosing a solution
and managing its costs and risks. Society may be continuously
informed and, if desired, build preparedness plans for critical
events.
Taking the maximum budget allowance for risks/emissions
mitigation equal to US 40 million, a compromise solution may
combine a moderate emission level say, 13 mm tCo2 and a
lower risk level say 1 GWmed expected load curtailment (in
case of a mandatory deficit). These parameters will lead to a
budget-constrained expansion, illustrated in Fig. 10.

[4]
[5]

[6]

[7]

[8]

[9]

[10]
[11]

[12]

[13]

[14]
Figure 10 Final expansion, emission- and risk-wise
[15]

R. Hemmati, R.-A. Hooshmand, A. Khodabakhshian, Comprehensive


review of generation and transmission expansion planning, IET
Generation, Transmission & Distribution, Volume 7, Issue 9, 2013, pp.
955-964
X. Wang, S. Xie, X. Wang, C. Qu, Decision-Making Model Based on
Conditional Risks and Conditional Costs in Power System Probabilistic
Planning, IEEE Transactions on Power Systems, Vol. 28, No. 4,
November 2013
A. J. Conejo, M Carrin, J. M. Morales, Decision Making under
Uncertainty in Electricity Markets, International Series in Operations
Research & Management Science, Springer; 2010
D. R. Biggar, M. R. Hesamzadeh, The Economics of Electricity
Markets, Wiley IEEE, 2014
Climate Change, the Smart Grid, and the Future of Electric Utilities
Smart Power Anniversary Edition (Paperback) - Common Paperback
2014 Publisher: Island Press (2014)
G.S. Svenda, S. Kanjuh, T. Konjic, V. Miranda, Using a Fuzzy
Modeling in Decision Making for Planning Under Uncertainty with
Risk Analysis Paradigm 8th Seminar on Neural Network Applications
in Electrical Engineering, 2006.
T.B. Lam, H.A. Abbass, M/Barlow, Robustness Against the DecisionMaker's Attitude to Risk in Problems With Conflicting Objectives,
IEEE Transactions on Evolutionary Computation, Vol.16, Iss 1 , 2012
L. Pinto, L.F.Nogueira, The oprimal expansion planning: the bennefits
of a renewable option, IEEE International Conference on Environment
and Electrical Engineering, Wroclau, Poland, 2013
L. Pinto, L.F.Nogueira, A renewable generation portfolio: maximizing
delivery, minimizing risks, International Conference on the European
Energy Market, Krakow, 2014
L Pinto, P. Leite, Smart-grid, green energy and responsive consumers:
a "smart-green" framework, IEEE Powertech 2011, Norway
Ministry of Sciency and Technology, Greenhouse gas inventory and
metholology, currently available at http://www.mct.gov.br/upd_blob/
0027/27596.pdf (in portuguese)
George B. Dantzig, Mukund N. Thapa, Linear Programming 1:
Introduction (Springer Series in Operations Research and Financial
Engineering) (v. 1), 1997, ISBN-13: 978-0387948331
Brazilian Planning Authority (EPE), Planning 2023 (PDE 2023),
currently available at http://www.epe.gov.br/Estudos/Documents/
PDE2023.pdf (in portuguese)
Brazilian Generation Information Data (BIG-ANEEL), currently
available
at
http://www.aneel.gov.br/aplicacoes/capacidadebrasil/
capacidadebrasil.cfm
J. Szczupak, S. Crombez, Long term climatological scenario
generation and uncertainty evaluation, IEEE International Conference
on Environment and Electrical Engineering, Krakow, Poland, 2014

You might also like