You are on page 1of 9

Quantifying self-consumption of on-site photovoltaic power

generation in households with electric vehicle home charging


Joakim Munkhammar
a,
, Pia Grahn
b
, Joakim Widen
a
a
Built Environment Energy Systems Group, Department of Engineering Sciences, Uppsala University, SE-751 21 Uppsala, Sweden
b
KTH Royal Institute of Technology, School of Electrical Engineering, Division of Electric Power Systems, Teknikringen 33, SE-100 44 Stockholm, Sweden
Received 23 November 2012; received in revised form 8 August 2013; accepted 15 August 2013
Available online 12 September 2013
Communicated by: Associate Editor Elias K. Stefanakos
Abstract
Photovoltaic (PV) power production and residential power demand are negatively correlated at high latitudes on both annual and
diurnal basis. If PV penetration levels increase, methods to deal with power overproduction in the local distribution grids are needed
to avoid costly grid reinforcements. Increased local consumption is one such option. The introduction of a home-charged plug-in electric
vehicle (PEV) has a signicant impact on the household load and potentially changes the coincidence between household load and pho-
tovoltaic power production. This paper uses a stochastic model to investigate the eect on the coincidence between household load and
photovoltaic power production when including a PEV load. The investigation is based on two system levels: (I) individual household
level and (II) aggregate household level. The stochastic model produces theoretical high-resolution load proles for household load
and home charged PEV load over time. The photovoltaic power production model is based on high-resolution irradiance data for Upp-
sala, Sweden. It is shown that the introduction of a PEV improves the self-consumption of the photovoltaic power both on an individual
and an aggregate level, but the increase is limited due to the low coincidence between the photovoltaic power production pattern and the
charging patterns of the PEV.
2013 Elsevier Ltd. All rights reserved.
Keywords: Plug-in electric vehicle; Distributed photovoltaics; Load matching; Self-consumption
1. Introduction
The rapid increase of installed PV capacity in a number
of countries over the last decade has been dependent on
support schemes that assign an advantageous price to the
produced electricity, such as feed-in taris. These support
schemes have also had the eect, however, that the system
owners own consumption has neither inuenced the pro-
duction value nor the PV system design. Without consider-
able feed-in taris the self-consumption of PV power is
economically advantageous since the price of the self-pro-
duced power, if sold on the electricity market, is lower
compared to the grid power due to taxes, grid feed, etc.
added to the end-user price. Self-consumption may also
be actively promoted with special rates within feed-in-tari
schemes, such as in Germany in 2009 (IEA-PVPS, 2011). A
major injection of unmatched PV power in the grid at the
end-user can also lead to voltage rise or capacity problems
(Walla, 2012; Widen, 2010). Increased self-consumption of
PV power could enable an increased amount of PV power
injected to the grid without the necessity for improving the
grid thus increasing the distribution grids so-called host-
ing capacity for PV. Other possible ways to increase the
hosting capacity include for example power curtailment,
reactive power supply and substation tap changer adjust-
ments (Braun, 2011; Walla, 2012). Self-consumption can
0038-092X/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.solener.2013.08.015

Corresponding author. Tel.: +46 704464271.


E-mail addresses: joakim.munkhammar@angstrom.uu.se (J. Mun-
khammar), joakim.widen@angstrom.uu.se (J. Widen).
www.elsevier.com/locate/solener
Available online at www.sciencedirect.com
ScienceDirect
Solar Energy 97 (2013) 208216
be increased with for example active and automated load
shifting, battery storage and thermal storage via for exam-
ple domestic hot water heating (Braun, 2011; Binder, 2012;
Castillo-Cagigal, 2010; IPVEA, 2011; SunEdison, 2011).
Automated systems for energy management exist on the
market, where storage of a fraction of the daily PV produc-
tion is performed and advanced surveillance of production
and load is possible, SMAs Sunny Home Manager is an
example of that (Beister, 2012; SMA, 2012). Since a battery
is included in every PEV without any additional cost
other than for the vehicle itself it would be interesting
to investigate the use of its battery as PV energy storage.
In order to estimate how much home charging can contrib-
ute to increased self-consumption we have to investigate
how household power consumption and PV power produc-
tion coincide and how this changes with the introduction of
a home-charged PEV. There exist a number of investiga-
tions regarding the grid impact of PEV load, in particular
recently (Haghbin, 2013; Steen, 2012, 2011). There are also
a number of investigations regarding PEV load and photo-
voltaic power production, such as studies on vehicle to grid
(V2G) technology (Cvetkovic, 2009; Kempton and Tomic,
2005) and optimization techniques (Guillou, 2011; Kinosh-
ita, 2011). A recent study investigates the benet of using
PV as a curtailment of PHEV load on a large scale (Den-
holm, in press). This investigation aims at providing an
analysis of how synthetic power consumption patterns of
residential power demand along with PEV use could be
used to estimate the coincidence with PV power production
at the end-user.
1.1. Aim of the study
The aim of this paper is to investigate the coincidence
between PV power and PEV charging and to study on
both an individual and aggregate level how the self-con-
sumption of PV in households is changed with the intro-
duction of home charged PEV load. The investigation is
made with a Markov-chain model for generating synthetic
household energy use patterns and PEV charging pat-
terns. The PV power production is given from high reso-
lution data of incident solar radiation from Uppsala,
Sweden.
1.2. Outline of the paper
In Section 2 the model and data are explained. In Sec-
tion 3 the results from simulations of the model are given.
In Section 4 the results are discussed.
2. Methodology
This paper utilizes a stochastic Markov-chain model
based on time-use data (TUD) for simulating household
electricity use and PEV charging. Resulting local load
and production proles are used together with simulated
PV generation proles to quantify self-consumption.
2.1. Energy consumption from activities
Time-use data regarding domestic activities over time is
useful in order to quantify domestic energy use (Capasso,
1994; Widen, 2010, 2012; Widen and Wackelgard, 2010).
Time-use data can for example be obtained via interviews
or time-diaries. The typical assumption is that each activity
performed by an individual has a certain amount of energy
use associated with it (Widen, 2012). This type of model
can then be validated by comparing the output of the
model to the high-resolution power consumption data
(Widen, 2010). As an alternative to developing determinis-
tic models based on time-use data it is possible to use that
data to calibrate a stochastic model which in turn can
produce synthetic activity patterns and consequent energy
use (Widen, 2009; Widen and Wackelgard, 2010). The
model for estimating energy consumption which is used
in this paper is based on a discrete-time stochastic Mar-
kov-chain model for generating synthetic activity data
(Widen, 2009).
2.2. Markov-chain synthetic activity generation
In Widen (2009) and Widen and Wackelgard (2010) a
Markov-chain model for generating synthetic activity data
over time was developed. This model will in the text be
referred to as the Markov-chain model or the stochastic
model. The model was based on the following two
principles:
(I) Each individual occupies one state of activity at each
time-step.
(II) The number of possible states is a xed number.
Based on these assumptions the Markov-chain model
assigns a probability for transition from one activity to
another at each time step, this includes the probability
for staying in the activity occupied during the previous
time-step. The transition probabilities in this model were
estimated from time-use activity data (Widen, 2009; Widen
and Wackelgard, 2010). Formally a discrete-time Markov
chain S
t
is a discrete stochastic process based on the pre-
mise the model occupies only one state E
l
in a number
of states dened by l 2 [1, . . . , N] for each time step. The
probability for each activity to become occupied in the next
state is then calculated from on the basis of the transition
matrix Q
lm
(t) (for time-step t 2 [1, . . . , T]):
Q
lm
t ProbX
t1
E
m
jX
t
E
l
: 1
The transition matrix is dened for l 2 [1, . . . , N] and
m 2 [1, . . . , N] where N is the number of activities. The prob-
ability for transition from one activity to another is only
dependent on the previous time-step; this property needs
to be satised for Markov-chains and is called the Markov
property. The transition matrix in Eq. (1) satises the Mar-
kov property since the state at t + 1 is only dependent on
J. Munkhammar et al. / Solar Energy 97 (2013) 208216 209
the state at t. From this it is possible to give an expression
for the probability of occupying a particular state E
l
at
time step t (Cinlar, 1975, p. 107):
p
l
t ProbX
t
E
l
: 2
Due to the rst principle in the list Section 2.2 the prob-
ability for occupying any of the predened activities in the
list has to equal one for each time step t:
X
N
l1
p
l
t 1: 3
In Table 1 the dierent states of the stochastic model is
given.
There are several dierent possibilities to construct load
proles from activities, where assumptions regarding appli-
ances have to be made. This model uses the assumptions
dened in Widen (2009) and Widen and Wackelgard
(2010). The stochastic model produces a binary activity
matrix A
j,t
where t 2 [1, . . . , T] is the time step and
j 2 [1, . . . , N] is the activity. If for some specic t, j the bin-
ary entry A
j,t
equals 1 then the particular activity j is active
at the particular time t. If A
j,t
on the other hand is 0 then
another activity than the specic activity j is active at that
time. The model has options to either model apartments or
detached houses. For this investigation only the detached
house option is used. The activity sequence generated in
the model is minute-based but the transition probabilities
are hourly averages to provide enough data for making
the estimates. For more detailed information regarding
the assumptions and the setup of the stochastic model see
(Widen and Wackelgard, 2010).
2.3. PEV home charging model
In this section the PEV home charging model is
described. The model is an extension of the stochastic
model described in Section 2.2. The PEV home charging
model is described in detail with various settings in Grahn
(in press). The variables that are used in the model are
listed in Table 2.
The PEV consumes electricity from the battery when it
is driven. The main assumption of the extension is that
the PEV is used with a probability fraction p
PEV
when
the inhabitant enters state away in the stochastic model.
Formally this is equivalent to when A
1,t
= 0 is changed to
A
1,t+1
= 1 (see Table 1 for the set of states of the stochastic
model). This is modeled with a parameter k which is a sto-
chastic variable k 2 U(0, 1) which randomized each time
the choice to take a car or not is present. The PEV is used
or not used during all time-steps for which A
1,t
= 1, thus k
is xed during those time-steps. Then when A
1,t
changes
from 1 to 0 and then later changed back to 1 again the ran-
dom parameter k is once again randomized and the process
whether the PEV is taken or not is evaluated. The PEV
home charging occurs when the inhabitant returns home
after a trip with the PEV. The arrival at home changes
the activity state A
1,t
= 0 for that particular time-step. Thus
the starting time and the returning time of each PEV trip is
decided by the synthetic activity pattern that is produced
by the stochastic model. When the PEV is parked at home
it is assumed that it is immediately plugged in and charged.
The charging takes place until the PEV is fully charged or it
is used again. For all calculations below the index t repre-
sents the time step t = 1, . . . , T where T is the total number
of time steps. The charging power when the PEV battery is
connected and charged is C
Charge
. The PEV charging load
P
PEV
(t) is then calculated as:
P
PEV
t
C
Charge
if charging;
0 else:
(
4
It is possible to construct a model for electric power ex-
ibility by letting the PEV discharge and thus produce elec-
tric power to the household. This possibility is however not
implemented in this model and the PEV is only considered
an electric power load. The grid-to-vehicle power ow is
thus P
PEV
(t) P0. To avoid reducing the lifetime of the bat-
tery the state of charge is only allowed to fall to a certain
depth of discharge (DOD) limit, here dened as a minimum
state of charge SOC
min
. The state of charge SOC(t) thus
has the following constrains:
SOC
min
< SOCt 6 SOC
max
: 5
The amount of electric power the PEV consumes during
its trip depends on the average driving electricity consump-
tion C
PEV
, multiplied by the seasonal factor S(t). At the
Table 1
Activity states.
Code Activity
1 Away
2 Sleeping
3 Cooking
4 Dishwashing
5 Washing
6 TV
7 Computer
8 Audio
9 Other
Table 2
Variables.
Denotation Variable Unit
Charging power C
Charge
kW
Load P
PEV
(t) kW
State of charge SOC(t) kW h
Maximum state of charge SOC
max
kW h
Minimum state of charge SOC
min
kW h
Electricity consumption C
PEV
kW
Vehicle usage probability p
PEV
Season coecient S(t)
210 J. Munkhammar et al. / Solar Energy 97 (2013) 208216
initial time-step t = 0 in the model the battery is initially
fully charged with SOC(0) = SOC
max
. The electric energy
level in the battery SOC(t) is calculated according to:
SOCt 1
SOCt C
PEV
StDt if consuming;
SOCt C
Charge
Dt if charging;
SOCt else:
8
>
<
>
:
6
One assumption that was made here was that if the PEV
was away for a longer period than the battery lasted (since
in practice the maximum time of use for the PEV without
charging is proportional to driving electricity consumption)
the PEV was assumed to have stopped during some part of
the trip and then depleted the battery upon arrival at home
(Grahn, 2011). It is also possible to assume that during
these trips the PEV had some other source of energy than
the battery, but that it just like the previous situation had
depleted the battery upon arrival at home.
In Grahn (2011) a standard setup was developed, which
had a setup according to Table 3.
The seasonal variation in fuel consumption for the stan-
dard setup was assumed to be a factor S(t) times the aver-
age fuel consumption. For the dierent seasons the
assumption in Grahn (in press) was according to the fol-
lowing. Winter: S(t) = 1.2, spring: S(t) = 1, summer:
S(t) = 0.8, fall: S(t) = 1.0. This coecient is intended to
mimic the variability of load from seasonal conditions
from for example heating.
2.4. PV data and model
The PV power output was simulated from high-resolu-
tion incident in-plane solar irradiance data. A simple
straight forward method for calculating the PV power out-
put was used in this paper. The power output of the PV
module is calculated from the formula:
P
PV
t g A Gt; 7
where P
PV
(t) is the power output over time, A is the PV
area (m
2
), g is the eciency of the PV system and G(t) is
the incident solar radiation (W/m
2
). The system eciency
was set to g = 13%. The incident solar radiation data
G(t) was 2-s resolution and measured in a plane tilted 45
with a pyranometer at the A

ngstrom laboratory at Uppsala


Sweden (5950
0
19
00
N 1738
0
50
00
E) during January 1Decem-
ber 31 2011. The data was averaged to 1 min resolution in
order to match the resolution of the household- and PEV
load from the stochastic model. The data was adjusted
for daylight saving time.
2.5. Denitions and measures
There are two useful measures of self-consumption that
will be used in this paper: Solar fraction of power demand
(SF) and Load fraction of the PV production (LF).
Since SF is dened as the fraction of load that is
matched with PV power it can mathematically be described
using the labels in the illustration in Fig. 1:
SF
B
A B C
: 8
Also since LF is dened as the fraction of PV power
which is matched by load it can mathematically be
described using the labels in the illustration in Fig. 1:
LF
B
B D
: 9
It should be noted that for net-zero energy buildings the
solar fraction equals the load fraction since the parts
A + B + C and B + D are equal if summed up over a whole
year.
2.6. Scenarios
The simulations made in this paper were based on the
following scenarios:
A. A single detached house with two inhabitants and no
PEV load.
B. A single detached house with two inhabitants where
one of the inhabitants uses the PEV.
C. An aggregate of 1000 detached houses with two
inhabitants in each household and no PEV load.
D. An aggregate of 1000 detached houses with two
inhabitants, where in each household one of the
inhabitants uses the PEV.
Table 3
Variables for the standard sce-
nario of the PEV.
C
Charge
2.3 kW
SOC
max
35 kW h
SOC
min
21 kW h
p
PEV
20%
S(t) 0.81.2
C
PEV
8.4 kW
Fig. 1. A schematic illustration of a load curve (solid) and production
curve (dashed) with labels for energy that is used to estimate solar fraction
and load fraction.
J. Munkhammar et al. / Solar Energy 97 (2013) 208216 211
For scenario A and B there are corresponding approxi-
mate yearly net-zero energy PV setup of PV1: A = 25 m
2
and PV2: A = 34 m
2
. One reason for investigating net-zero
energy scenarios are that both systems (with or without
PEV load) are net-zero on a yearly basis which makes
the size of the PV systems are comparable so that self-con-
sumption measures such as solar fraction and load fraction
more accurately may be compared. Assuming a module
eciency of about 17% this corresponds to the following
peak power: PV1: 4.3 kW
p
and PV2: 5.8 kW
p
. Due to sim-
ilarity in load for each household the same PV setup as for
scenario A and B is used for each household in scenarios C
and D.
3. Results
In order to properly quantify how the introduction of
PEV charging to the household load aects the coincidence
between load and PV power production the scenarios in
Section 2.6 were investigated. The setup on the PEV in
the simulations is the standard setup developed in Grahn
(2011, in press) and found in Table 3. The rst simulation is
of scenario A. The output from the simulation is household
load including the load from the home charging of the
PEV. Fig. 2 shows the average household divided into
the dierent categories of appliances.
The PEV load is about 1.53 MW h/year and the total
household load including PEV load is about 5.7 MW h/
year which makes the PEV load about 27% of total house-
hold load. In a household with fewer than two inhabitants
the number of inhabitants in this simulation the frac-
tion of PEV load would be even higher. In Table 4 annual
energy consumption/production, max power and standard
deviation of power for the dierent components of the
investigations in the scenarios are given.
The variability of both the household load including
PEV and the PV power production is shown in the four
day sample of PV power output and household load
including PEV load in Fig. 3. In Fig. 4 the average output
from both PV setup and household with or without PEV
load on both individual and aggregate level is shown. In
that gure it is possible to notice that the PEV has a con-
siderable contribution to the household load, mainly dur-
ing late evening, night time and morning (Grahn, in
press).
Table 5 shows results regarding self-consumption mea-
sures for the dierent scenarios. In both the PV1 (25 m
2
)
and PV2 (34 m
2
PV) array sizes the standard deviation of
load is increased when the PEV is introduced. In the indi-
vidual scenario the standard deviation is more than dou-
bled, in the aggregate scenario the standard deviation is
almost doubled, see Tables 46. The reason for this is the
strong intermittency of the PEV load. especially on single
household level. The introduction of the PEV does not
have any particular eect on the standard deviation of
the production though. In the same tables the maximum
net load when comparing the cases of household including
PEV and household including PEV and PV array is not
changed. The lack of change is probably related to the fact
that maximum load occurs during late evening when PV
power production is low or zero. The maximum PV power
production on the other hand is changed when introducing
PV, although the change is only on the order of a few per-
cent in both individual and aggregate scenarios. A house-
hold without a PEV is about net-zero with a 25 m
2
PV
array setup. It is interesting to note that if a PEV is added
to this setup the solar fraction decreases from about 32% to
about 26%, while the load fraction increases from about
31% to about 34% and the total energy consumption of
the household is increased by around 37%. If the PV array
0 5 10 15 20
0
0.2
0.4
0.6
0.8
1
1.2
Time (Hours)
P
o
w
e
r

(
k
W
)
A: Cold Appliances
B: Cooking
C: Washing
D: Dishwashing
E: Television
F: Computer
G: Audio
H: Lighting
I: Additional Appliances
J: Electric Vehicle
Fig. 2. A stacked bar plot over the activities and their power use averaged
over time.
Table 4
Load, PV power production and standard deviation for the dierent scenarios over a year.
Load/production unit Annual energy (MW h/
household)
Max. power (kW/
household)
Std power (kW/
household)
Household 4.17 4.52 0.37
Household + PEV 5.7 6.25 0.73
1000 Households 4.15 0.85 0.12
1000
Households + PEV
5.83 1.26 0.25
PV1 (25 m
2
) 4.21 4.56 0.87
PV2 (34 m
2
) 5.73 6.20 1.18
212 J. Munkhammar et al. / Solar Energy 97 (2013) 208216
size is changed to 34 m
2
in order to match the extra PEV
power then the solar fraction is increased from about
26% to 31% which is lower than the solar fraction in
the rst case with the household without PEV load with
corresponding net-zero setup. Thus the household includ-
ing PEV and corresponding net-zero energy PV setup
yields lower solar fraction than a household without PEV
with corresponding net-zero energy setup. Given the setup
of a household without PEV load with plus energy setup of
having the PV size adapted to give net-zero energy setup
for household including PEV load, then it is worth noting
that the load fraction of the PV-load increases from about
25% to about 28% in the single household scenario and it
increases from about 27% to about 32% in the aggregate
scenario. It is important to point out that a net-zero energy
building with a PEV has lower self-consumption than a
1 2 3 4
0
1
2
3
4
5
Time (Days)
P
o
w
e
r

(
k
W
)
PEV+Household
PV2
Fig. 3. An example plot of household load including EV load in comparison with photovoltaic power production over four days in April 2011.
00:00 12:00 24:00
0
0.5
1
1.5
2
Individual household
Time (Hours)
P
o
w
e
r

(
k
W
)
00:00 12:00 24:00
0
0.5
1
1.5
2
Aggregate of households
Time (Hours)
P
o
w
e
r

(
k
W
)
EV+Household
Household
34m
2
PV
25m
2
PV
Fig. 4. Year-average daily power use. On the left hand side the single household with and without PEV load for the two options of PV-setup of 25 m
2
and
34 m
2
is shown. On the right hand side the year average of an aggregate of 1000 households is presented for both the scenario with and without a PEV.
Table 5
Power and standard deviation of power for the individual household scenarios for various setup. PV1 = PV power production from 25 m
2
, PV2 = PV
power production from 34 m
2
, H = Household load, PEV = PEV load.
Load/production
unit
Solar fraction of
load (%)
Load fraction of
solar (%)
Std of net production
(kW)
Std of net load
(kW)
Max net production
(kW)
Max net load
(kW)
PV1-H 31.6 31.3 0.96 0.37 4.27 4.52
PV1-H-PEV 25.5 34.4 0.96 0.76 4.23 6.25
PV2-H 34.4 25.0 1.33 0.37 5.91 4.52
PV2-H-PEV 28.0 28.0 1.33 0.76 5.87 6.25
Table 6
Power and standard deviation of power for the aggregate scenarios with 1000 households for dierent type of setup. PV1 = PV power production from
25 m
2
, PV2 = PV power production from 34 m
2
, H = Household load, PEV = PEV load.
Load/
production unit
Solar fraction of
load (%)
Load fraction of
solar (%)
Std of net production
(kW)
Std of net load
(kW)
Max net production
(kW)
Max net load
(kW)
PV1-H 34.2 33.7 0.93 0.16 4.13 0.85
PV1-H-PEV 29.1 40.2 0.89 0.30 4.04 1.26
PV2-H 36.8 26.7 1.31 0.16 5.77 0.85
PV2-H-PEV 31.7 32.2 1.28 0.30 5.68 1.26
J. Munkhammar et al. / Solar Energy 97 (2013) 208216 213
net-zero energy building without, which is related to the
low coincidence between PEV electricity consumption
and PV power production.
A comparison can then be made between the results in
the table for the results from the individual household sce-
narios Table 5 and the 1000 household aggregate scenario
Table 6. In all cases the aggregate scenarios have higher
solar fraction and the same goes for load fraction. By any
means of measure the aggregate scenario has a more signif-
icant coincidence between PV power production and load.
It is worth noticing that the dierence between standard
deviation of net production is not large between individual
household level and aggregate level. This is an artefact of
the fact that the same PV data has been used on the average
of every household in the aggregate scenario. This is per-
haps not realistic since spatial distribution of PV panels
might give a less uctuant averaged PV power output. This
aspect although interesting is not included in this paper
for simplifying reasons. If one instead observes the standard
deviation of net consumption in Tables 5 and 6 the conclu-
sion is that it is invariably lower for all aggregate scenarios
compared with the corresponding individual scenarios. This
result stems from the convergence of the average output
from a large number of individual households where each
displays stochastic behavior. However in both the individ-
ual household scenarios and aggregate household scenarios
the standard deviation is higher for the scenarios where the
PEV is introduced. Note the near equality between solar
fraction and load fraction when the setup is net-zero energy
in both the individual and the aggregate scenarios in Tables
5 and 6. PV1-H and PV2-H-PEV net-zero energy
whereas PV1-H-PEV and PV2-H are not.
0 2000 4000 6000 8000
5
0
5
Hours
P
o
w
e
r

(
k
W
)
Aggregate of households
PV2
PV1
Household
Household+EV
0 2000 4000 6000 8000
5
0
5
Hours
P
o
w
e
r

(
k
W
)
Aggregate of households
Household+EVPV2
Household+EVPV1
HouseholdPV2
HouseholdPV1
0 2000 4000 6000 8000
5
0
5
Hours
P
o
w
e
r

(
k
W
)
Individual household
Household+EVPV2
Household+EVPV1
HouseholdPV2
HouseholdPV1
0 2000 4000 6000 8000
5
0
5
Hours
P
o
w
e
r

(
k
W
)
Individual household
PV2
PV1
Household
Household+EV
Fig. 5. This gure represents a duration plot over household load with included PEV load and the two scenarios with PV production: PV1 = 25 m
2
and
PV2 = 34 m
2
. On the left hand side the duration plot of a single household over a year is presented, on the right hand side the duration plot for an average
household in an aggregate of 1000 households is presented (also over a year). The result above zero is the magnitude of electricity consumption, below zero
is a measure of the magnitude of electricity production.
0 10 20 30 40
0
1
2
3
4
5
6
7
PV array area (m)
A
v
e
r
a
g
e

e
n
e
r
g
y

(
M
W
h
/
y
e
a
r
)
Single household
0 10 20 30 40
0
1
2
3
4
5
6
7
PV array area (m)
A
v
e
r
a
g
e

e
n
e
r
g
y

(
M
W
h
/
y
e
a
r
)
1000 Households
Matched Household+EV
Matched Household
PV production
Household+EV
Household
Fig. 6. The amount of matched PV production for the dierent scenarios of including PEV load or not. The single household scenarios are on the left hand
side and the 1000 household aggregate is on the right hand side.
214 J. Munkhammar et al. / Solar Energy 97 (2013) 208216
In order to illustrate the variability between electricity
consumption and PV power production a duration plot
was made for both the individual scenarios A and B and
the aggregate scenarios C and D, see Fig. 5. The gure
shows the duration of the individual load and production
from household load and PV production separated as well
as the total load/production from the household where all
is included. For both the individual household scenarios
and aggregate household scenarios it is apparent in this g-
ure that there is a similarity between scenarios with the
same PV setup. One observation is that the dierence in
the duration plot between dierent PV scenarios is larger
than the dierence between adding PEV or not. It is also
possible to notice that the aggregate scenarios also have
considerably lower peaks in consumption.
All investigations so far have regarded two xed PV
panel sizes that are equivalent to net-zero energy for house-
holds including PEV load or not respectively: PV2 = 34 m
2
and PV1 = 25 m
2
. As a generalization of this concept the
solar fraction for a wide range of PV sizes from zero up
to 40 m
2
is given in Fig. 6. As PV panel size increases the
level of PV power that is matched by household electricity
consumption with or without PEV load (SF) increases, but
the fraction of the amount that is matched relative to the
PV production decreases (LF).
4. Discussion
The introduction of a home-charged PEV has conse-
quences for the load prole of the household with a local
PV system. These consequences are dierent on individual
and an aggregate level. In this paper the following four sce-
narios were investigated: (A) A single household without a
PEV, (B) a single household with a PEV, (C) an aggregate
of 1000 households without a PEV, and (D) an aggregate
of 1000 households each with a PEV. These scenarios were
equipped with two PV panel sizes: PV1 = 25 m
2
and
PV2 = 34 m
2
which were based on net-zero energy of sce-
narios (C) and (D) on a yearly basis. Simulations show that
a PEV gives a considerable contribution to the household
load and that most of a PEV load is concentrated during
late evenings, night time and early mornings. Together with
a highly intermittent and high peak power consumption a
PEV alters the coincidence between the household load
and the local PV power. For the single household scenarios
the solar fraction is decreased when introducing a PEV
load. This is true on both the individual household level
and an aggregate of households. However results show that
all aggregate scenarios have higher solar fraction than their
individual scenario counterpart. The introduction of PEV
load increases the load fraction both on individual and
aggregate levels. However the introduction of a PEV in a
local net zero energy setup but with increased PV size (as
to make household load including PEV load net-zero)
makes both solar fraction and load fraction lower than in
the situation with net zero energy setup of a household
without PEV load. This stems from the fact that the level
of coincidence between PEV load and PV electricity pro-
duction is lower than the coincidence between household
load an PV electricity production. It should be noted here
that we have not considered any limit on the maximum
potential household PV array area a factor necessary to
consider in for example case studies involving sizing up
PV for a particular building in order to achieve optimal
matching between PEV load and PV power production.
Results also show that an aggregate scenario would be
more benecial for increasing self-consumption of PV
power. This is an artefact from the signicantly lower stan-
dard deviation of load in the aggregate scenarios compared
with the individual scenarios. The default home-charging
studied in this paper is not optimal if maximized utilization
of PV power is the aim. If PEV charging at work was
enabled then the coincidence between the charging power
and the PV power production could presumably be
increased since normal work hours usually coincide with
daylight.
5. Conclusions
The introduction of a home-charged PEV increases the
household load from 4.17 MW h/year to 5.7 MW h/year
which is an increase of 37%. The introduction of PEV
load to a household with net zero energy setup makes
the solar fraction drop 20% while the load fraction
increases by 10%. If a household without PEV has net-
zero energy setup and a PEV along with additional PV
area is added to the house so that the house once again
is net-zero energy then the solar fraction increases with
13% and the load fraction decreases by 10%. While hav-
ing about equal LF and SF in each of both net-energy
scenarios their magnitude is lower for the net-zero energy
household with PEV due to the higher mismatch between
load and production.
Acknowledgements
I would like to thank Uwe Zimmermann at the Depart-
ment of Engineering Sciences at Uppsala University for
providing the high-resolution data for the PV-simulations.
This work has been carried out under the auspices of The
Energy Systems Programme, which is primarily nanced
by the Swedish Energy Agency.
References
Beister, D., 2012. Sophisticated energy management for optimizing the PV
grid interaction. In: Conference Proceeding 27th EU-PVSEC, pp.
41014104.
Binder, J. et al., 2012. Increasing PV self-consumption, domestic energy
autonomy and grid compatibility of PV systems using heat-pumps,
thermal storage and battery storage. In: Proceedings of 27:th EU-
PVSEC.
Braun, M. et al., 2011. Is the distribution grid ready to accept large scale
photovoltaic deployment? State of the ARt, progress and future
prospects. In: Conference Proceeding 26th EU-PVSEC, pp. 3840
3853.
J. Munkhammar et al. / Solar Energy 97 (2013) 208216 215
Capasso, A. et al., 1994. A bottom-up approach to residential load
modeling. IEEE Transactions on Power Systems 9 (2), 95796.
Castillo-Cagigal, M. et al., 2010. Self-consumption of PV electricity with
active demand side management: the GEDELOS-PV system. In:
Conference Proceeding 25th EU-PVSEC, pp. 48664870.
Cinlar, E., 1975. Introduction to Stochastic Processes. Prentice-Hall Inc..
Cvetkovic, I. et al., 2009. Future home uninterruptible renewable energy
system with vehicle-to-grid technology. Conference Proceeding ECCE.
IEEE.
Denholm, P. et al., in press. Co-benets of large scale plug-in hybrid
electric vehicle and solar PV deployment. Journal of Power Sources (in
press).
Grahn, P. et al., 2011. Photovoltaics, Electric Vehicles and Energy Users:
A Case Study of the Royal Seaport, Visions and Energy User
Expectations. Working Paper No. 50 at Linko ping University.
Grahn, P. et al., in press. Plug-in-vehicle home charging model based on
residential activity patterns. IEEE Transactions of Power Systems (in
press).
Guillou, H. et al., 2011. Energy management strategies for optimal
charging of electric vehicles with photovoltaic production. In:
Conference Proceeding PVSEC, pp. 38853889.
Haghbin, S. et al., 2013. Grid-connected integrated battery chargers in
vehicle applications: review and new solution. IEEE Transactions on
Industrial Electronics 60 (2), 459472.
IEA-PVPS, 2011. Germany National Photovoltaics Survey Report,
Report.
IPVEA, 2011. Super charged: self-consumption sparks alliances between
PV and storage industries. Energy Storage Journal (1).
Kempton, W., Tomic, J., 2005. Vehicle-to-grid power implementation:
from stabilizing the grid to supporting large-scale renewable energy.
Journal of Power Sources 144 (1), 280294.
Kinoshita, S. et al., 2011. Operation design of a PV system with storage
battery for EV quick charger at small retail store. In: Conference
Proceedings PVSEC.
SMA, Sunny Home Manager. <http://www.sma.de/en/produkte/moni-
toring-systems/sunny-home-manager.html> (07.11.12).
Steen, D. et al., 2011. Optimal load management of electric heating and
PEV loads in a residential distribution system in Sweden. In:
Proceedings IEEE PES Innovative Smart Grid Technologies Confer-
ence, pp. 16.
Steen, D. et al., 2012. Assessment of electric vehicle charging scenarios
based on demographical data. IEEE Transactions on Smart Grid 3 (3),
14571468.
SunEdison, 2011. Enabling the European Consumer to Generate Power
for Self-consumption, SunEdison Report.
Walla, T. et al., 2012. Determining and increasing the hosting capacity for
photovoltaics in swedish distribution grids. In: Conference Proceed-
ings 27th EU-PVSection, pp. 44144420.
Widen, J. et al., 2009. combined Markov-chain and bottom-up approach
to modelling of domestic lighting demand. Energy in Buildings 41 (10),
10011012.
Widen, J. et al., 2010. Constructing load proles for household electricity
and hot water use from time-use data modelling approach and
validation. Energy and Buildings 41 (7), 753768.
Widen, J. et al., 2010. Impacts of distributed photovoltaics on network
voltages: stochastic simulations of three Swedish low-voltage distribu-
tion grids. Electric Power Systems Research 80 (12), 15621571.
Widen, J. et al., 2012. Models of domestic occupancy, activities and energy
use based on time-use data: deterministic and stochastic approaches
with application to various building-related simulations. Journal of
Building Performance Simulations 5 (1), 2744.
Widen, J., Wackelgard, E., 2010. A high-resolution stochastic model of
domestic activity patterns and electricity demand. Applied Energy 87
(6), 18801892.
216 J. Munkhammar et al. / Solar Energy 97 (2013) 208216

You might also like