You are on page 1of 9

Renewable Energy 68 (2014) 443e451

Contents lists available at ScienceDirect

Renewable Energy
journal homepage:

Photovoltaics (PV) and electric vehicle-to-grid (V2G) strategies for

peak demand reduction in urban regions in Brazil in a smart grid
Lukas Drude a, b, Luiz Carlos Pereira Junior a, Ricardo Rüther a, *
Universidade Federal de Santa Catarina e UFSC, Caixa Postal 476, Florianópolis-SC 88040-900, Brazil
Universität Paderborn, Warburger Straße 100, Paderborn D-33098, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Vehicle-to-grid (V2G) energy transfer in a smart grid environment opens a new revenue opportunity for
Received 1 May 2013 electric-drive vehicles (EVs), and might reduce grid operation costs in demand-constrained urban
Accepted 30 January 2014 feeders where peak-electricity prices are high. This paper analyses the peak demand energy market for
Available online 12 March 2014
V2G in the urban region of Florianópolis, Brazil. The article describes known V2G-concepts and in-
troduces two different dispatch strategies developed for the Brazilian energy market in the light of new
tariff regulations, which are expected to go into effect starting in 2014. It turns out that electric-drive
vehicles may be used as a grid-stabilisation strategy, but the announced tariff regulations may lead to
Vehicle-to-grid (V2G)
Electric vehicles
a destabilisation if there are too many cars offering their internal storage for V2G grid support. Adequate
energy policy strategies must be introduced to avoid the conflicting interests that might stem from the
different perspectives of grid operators and EV owners. In contrast to previous analyses, this research
assumes the battery degradation as a function of depth of discharge (DOD), which is known for a specific
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Many countries have supported renewable energies with public

funds aiming for reduction of carbon emissions and the greenhouse
The traditional utility concept relies on a relatively small num- effect. The reduction of carbon emissions in the transport section,
ber of fairly large and centralised power plants, which quite often which is mainly based on internal combustion engines, is still a big
are distant from the urban centres where energy is consumed. In a challenge. Because of this, electric vehicles are frequently
large country like Brazil (8.5 million km2), transmission and dis- mentioned as a solution to the dependency on fossil fuels in the
tribution infrastructure and associated losses are considerable. transport sector [2]. Electric vehicles (EVs) have been experiencing
Hydropower generation plays a fundamental role in Brazil. considerable development in recent years [3,4], and pure plug-in
Growing environmental restrictions, and the larger distances from battery vehicles are now commercially available from a number
urban centres to the remaining potential, however, will lead to an of car manufacturers worldwide.
increase in end-user prices of electricity from new hydropower EVs have been proposed as a new power source for electric
plants. In contrast to this, the majority of the countries of the world utilities in the US and Japan in the late 90s [5,6]. Vehicle-to-grid
depends to a great extend on fossil energy sources like coal, pe- (V2G) power uses electric-drive vehicles to provide power for
troleum, and natural gas [1]. The Brazilian interconnected elec- specific electricity markets, since the electric power grid and light
tricity system is one of the largest and most complex in the world, vehicle fleets are exceptionally complementary as systems for
with an installed capacity close to 120 GW [1]. managing energy and power [7].
In a smart grid environment, a control signal from the grid
operator can send a request for power to a large number of parked
and plugged-in EVs to feed energy to the distribution grid. Yan and
* Corresponding author. Tel.: þ55 48 3721 5174; fax: þ55 48 3721 7615.
Paterson [8] have reported on power management systems for high
E-mail addresses:,, ruther@mbox1. (R. Rüther). performance and cost reduction in an EV, and as information
URL: technology and battery technology evolve, the opportunities for
0960-1481/Ó 2014 Elsevier Ltd. All rights reserved.
444 L. Drude et al. / Renewable Energy 68 (2014) 443e451

EVs to assist the public electricity grid in the V2G concept and in a demand, and defer investments in new power plants and trans-
smart grid environment become ever more real. With the devel- mission lines. This new tariff model will be optional for all resi-
opment of batteries that can operate satisfactorily for up to 7000 dential consumers. To choose the new model, the installation of
cycles1 (close to 20 years at one cycle per day), the additional digital electricity meters is necessary, to display information about
cycling of batteries due to V2G becomes compatible with the ex- energy consumption separately for each demand period [24]. These
pected life cycle of an EV. smarter meters will also enable the establishment of a smart grid in
Vehicle-to-grid energy transfer has already been analysed in the country. In this context, our paper presents a novel and original
different countries to modulate varying demands or unpredicted contribution in analysing the effects and potential contribution that
demand or supply availability variations. Ekman has studied the EVs and V2G can offer to the electricity sector in Brazil.
synergy between large EV fleets and high wind penetration in
Denmark [9]. There is a variety of analyses for the US energy market 2. Method
[7,10e13] Japan [6,14], Australia [15], and also the European grid
[16e20], but up to now there is no published research on EV To analyse the impact of V2G in the urban area under study, a
integration and effects in the context of the distribution network MATLAB simulation environment was implemented. This envi-
and the particular demand curve profile in the ever growing Bra- ronment allows the import of measured demand data, allows the
zilian interconnected electricity system. usage of different charge and discharge strategies, and the anal-
Electric-drive vehicles for V2G can be either hybrid, fuel cell, or ysis of the influence of the V2G-interactions on the given energy
pure battery vehicles. They are all EVs in the sense that they all use demand profiles. The simulation delivers the possibility to
an electric motor to provide all or part of the mechanical drive calculate annual costs of battery degradation, energy costs and
power. Battery-drive vehicles, fuel cell and hybrid drive trains have revenue in relation to parameters like maximum depth of
been analysed for the different energy markets base-load, peak discharge (DOD).
load, spinning-reserves and regulation-services [21]. Battery-drive
vehicles are highly suitable to supplement spinning-reserves and 2.1. Demand data and photovoltaic generation potential
regulation-services, as grid connection is already paid for, and in
this paper we analyse battery-drive vehicles for the peak-load In the simulation presented in this study, real demand data were
market in Brazil. used, which were measured in the urban residential area supplied
To enable V2G, each individual EV must meet three required by the #TDE05 feeder in the city of Florianópolis, Brazil in 2009. The
elements: (i) a connection to the grid for electrical energy flow, (ii) area consists of 6949 residential houses and 1099 commercial
control or logical connection to allow for communication with the buildings [25]. Demand in this area is measured at 15-min intervals.
grid operator, and (iii) a metering device on-board of the vehicle. There were 43 missing days in the demand data measured in
Pure plug-in battery EVs already have a grid connection to allow for 2009 and made available by the local distribution utility CELESC.
charging, and the incremental costs and operational adjustments to The missing days have not been considered in the calculation. To
add V2G are negligible [5,11]. To supply the electric motor, power correct this, all values like annual costs were extrapolated to a
electronics capable of generating grid voltage and frequency are whole year again. The original power demand values along the year
already built-in. Furthermore, battery EVs have larger batteries (e.g. are shown in Fig. 1. The (365e43) days of the year are displayed
24 kWh battery capacity for the Nissan Leaf) than hybrid EVs (e.g. along the horizontal axis, and the 24 h of the day are shown in the
1.6 kWh for the Toyota Camry) or plug-in hybrids (e.g. 6 kWh for vertical axis. Fig. 1 shows the demand profile for the year 2009,
the DaimlereChrysler Sprinter). Fuel cell and hybrid vehicles do not where the daily early evening peak demand is clearly seen all year
necessarily need a grid connection; thus they are subject to a higher round, and the mid-afternoon peak due to the extensive and
investment cost to allow for V2G. increasing use of air-conditioning units is seen only in summer
In most of the Brazilian urban centres, due to the extensive use months. V2G has thus an important role to play in assisting the
of electric resistance showerheads for water heating, energy de- reduction of the early evening peak, when drivers return to their
mand peaks in the early evening, from 19:00 to 21:00 [22,23]. homes and could plug their EVs to their residential power outlets to
Distribution utilities impose prohibitively high tariffs to commer- provide important support in assisting the grid from between
cial and industrial electricity consumers in this period to avoid even 18:00 and 19:00 up to 23:00 or 24:00. The daytime demand peak
higher power demands. These demand peaks coincide with the
typical urban driver weekday habits of returning home from a
working day. As from 2015, a new energy tariff structure is ex- 24 10
pected to come into effect in Brazil, and electricity will be charged 22
differently along the day even for residential consumers. For cus- 20
tomers choosing this option there will be differences according to 18
the peak of electricity demand. This new tariff model, known as 16
white tariff, will introduce three different payment levels or pe-
Time (hours)

Power (MW)

riods: peak demand, off-peak demand and intermediate demand.
This new rule defines a 5 higher tariff during peak demand than
10 4
during off-peak demand. Electricity in the intermediate demand
period will be charged 3 higher than during the off-peak demand 8

period. All three demand levels are fixed time periods along the 6
day, where peak demand is defined as the time period between 4
18:30 and 21:30 due to the high electric shower demands previ- 2
ously mentioned. The objective of this new tariff law is to induce 0 0
change in consumer habits, shift and reduce the peak power 50 100 150 200 250 300
Date (days)

Fig. 1. Original electricity demand (MW) profile of the mixed residential/commercial

1 area supplied by feeder #TDE05 in Florianópolis-Brazil in 2009.
L. Drude et al. / Renewable Energy 68 (2014) 443e451 445

that increasingly occurs in summer months can be reduced by the a low DOD will lead to very long battery lifetimes, measured in
use of building-integrated and grid-connected rooftop PV genera- cycles. This simulation is based on the approach of Rosenkranz and
tors, as we will show. the Fraunhofer ISI’s model, as shown in the following equation
To calculate the potential power that a number of rooftop [26,35]:
photovoltaic installations could supply in the urban area fed by the
#TDE05 feeder, which is a mixed urban region as previously NCycles ¼ 1331 DOD1:8248 (1)
mentioned, research previously carried out was considered [27].
The study evaluated the tilt angle and azimuthal orientation of roof where NCycles is the expected battery lifetime in cycles as plotted in
surfaces in this region to yield a good estimation of potential PV Fig. 2. This equation leads to an approximately linear dependence of
capacity of residential roofs. According to that study, more than 50% costs per stored energy, depending on the DOD, which can be seen
of the roof surfaces are suitable for the integration of PV modules at in Fig. 3.
tilt angles and orientations that lead to at least 95% of the The charging and discharging powers are limited by the grid
maximum possible power output. In this work, 230 W, 18.5% effi- connection. In Brazil most of the standard plugs allow a
cient multicrystalline silicon solar modules were evaluated, and maximum current of 10 A. For electrical showerheads and other
measured solar radiation data from year 2009 were used. larger electrical loads, a maximum current of 20 A is also common
Real PV performance data were obtained from the 2 kWp PV and available in all households. Thus we limit the maximum
generator that has been operating at the solar energy research power at PMax ¼ 4.4 kW, based on the standard 20 A sockets, at
laboratory at Universidade Federal de Santa Catarina, which were 220 V.
used in order to estimate the PV generation in our simulations. The The DCeAC converter efficiency is assumed as hconverter ¼ 98%,
comprehensive data acquisition system in that PV installation since state-of-the-art, commercially available inverters for photo-
measures ambient and PV module temperature, horizontal and voltaic applications reach this efficiency [36]. The efficiencies and
plane-of-array solar radiation, and electrical parameters at 5-min power limits are shown in the flow diagram in Fig. 4. In this figure
intervals since 1997 [28e30]. also the points for Pcharge and PDischarge are defined, which will be
further defined.
2.2. Assumptions
2.3. Charging
There are only a few studies regarding vehicle movement in
Brazil available. It was assumed that the electric-driven vehicles are Charging is performed between midnight and 6:00 with a
used during the day for shopping and working. It was also assumed charge rate as follows, where ESoC is the energy stored in the
that all cars are available for V2G as of 18:00, while being discon- battery:
nected from the grid during the day. The shortfall in electrical en- !
ergy represented by the energy needed to supply the fleet of EVs ECapa  ESoC ðnÞ
PCharge ðnÞ ¼ min ; hBattery PMax (2)
driving a given distance per year, is supplied by building-integrated tRemainingðnÞ
and building-applied photovoltaic (BIPV and BAPV) generators
spread on rooftops around the corresponding #TDE05 feeder area, where ESoC(n) is the energy at the given time-step, in kWh. The
as proposed by Santos and Rüther [27]. value ECapa is the battery capacity drawn from Table 1. The value
For price calculations the current energy price in Florianópolis is tRemaining(n) is calculated as the time difference between the current
US$ 0.1772/kWh, with a number of local taxes (PIS ¼ 1.65%; time-step and the end of the charge interval, measured in hours.
COFINS ¼ 7.6%; and ICMS ¼ 25%) already included [31]. The The new State of Charge (SoC) is therefore
assumed car specifications were drawn from Dallinger et al. [26] as
shown in Table 1. The battery price assumed from the international ESoC ðn þ 1Þ ¼ ESoC ðnÞ þ PCharge ðnÞDt (3)
market was US$ 300/kWh [32], with import tax (II ¼ 16%), federal
tax (IPI ¼ 15%) [33] as well as the above-mentioned local taxes PIS, with time-steps Dt ¼ 15 min.
COFINS and ICMS (12%). The demand seen by the grid is therefore increased to
The metering device (US$ 50), a communication device (US$ 50),
and the additional hardware needed to allow to feed energy to the PCharge ðnÞ
DNew ðnÞ ¼ DOld ðnÞ þ (4)
grid (US$ 33.33/kW) were calculated with the same local taxes, but hBattery hCharge
with the import tax II ¼ 14%, and the IPI tax ¼ 15%, following local
legislation [33].
All these values are summarised in Table 1, and are expected to
decrease with technology development and scaling effects. 10
Although the battery efficiency and lifetime both depend on the
charge and discharge rate, battery efficiency was assumed at 84%
[34]. It is known that the battery suffers from deep cycling, and that 5

Table 1
Electric vehicle assumed for the simulations in this study (data from Dal-
linger et al. [26], shown in both the Brazilian currency R$, and in US$). 10
Description Value

Battery capacity 25 kWh

Battery price R$ 10,201 $ 5481 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Meter for invoicing R$ 140 $ 75
Communication system R$ 280 $ 150
Additional electronics R$ 410 $ 220
Fig. 2. Estimated battery lifetime in cycles, as a function of the depth of discharge e
Maximum driving range 200 km
446 L. Drude et al. / Renewable Energy 68 (2014) 443e451

0.2 recharged in the early hours (G2V), and assist the utility grid during
Energy Costs (US$/kWh)

peak hours in the evenings (V2G).

Power trading like proposed by Hartmann and Özdemir [19] can
add additional revenue to EV car owners, but in most cases this has
been analysed with spot-market prices like the European Energy
0.1 Exchange (EEX). In Brazil energy is traded mostly in auctions, which
are held for long term contracts as described in Camargo et al. [39],
and Losekann and Oliveira [40], which are not targeting, and not
appropriate, to stabilise the grid directly.
In Strategy 1, batteries are charged in the morning and energy is
0 sold only during the peak hours interval, even if the actual demand
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
is already low by any reason. The power being discharged to the
grid, shown in Fig. 4, is estimated from the available energy and the
Fig. 3. Energy storage costs as a function of the depth of discharge e DOD. limits of power electronics:
where DNew is the resulting demand, and DOld is the original de- ESoc ðnÞ 1DODMax ECapa PMax
PDischarge ðnÞ¼min ; (6)
mand, as measured by the distribution utility company. tRemainingðnÞ hConverter
The parameter DODMax is the maximum allowed DOD, which is
2.4. Discharge caused by driving going to be varied during different simulations. Here PMax is divided
by the converter efficiency hConverter, because PDischarge is measured
Everyday, the energy for driving is discharged. The Brazilian before the converter, as defined in Fig. 4.
Interprofessional Group in Automotive Products and Services The battery charge level is updated according to Eq. (7):
calculated a medium daily car usage of 40 km in Brazil [37]. With
the maximum driving range given in Table 1 for the EV selected for ESoC ðn þ 1Þ ¼ ESoC ðnÞ  PDischarge ðnÞDt (7)
this study, the 200 km result in a driving energy EDrive ¼ 5 kWh for
The demand values are, as a consequence, reduced to:
average daily driving.
DNew ðnÞ ¼ DOld ðnÞ  hBattery PDischarge ðnÞ (8)
ESoC ðn þ 1Þ ¼ ESoC ðnÞ  EDrive (5)
While the analysis of daytime charging is outside the scope of
2.6. Strategy 2: grid stabilisation
the present work, in the medium term plugging in might also be
possible during the day for staff driving to work every day and
For a fixed number of 250 EVs, as proposed in a recent study
parking their cars in assigned parking lots in many public and
carried out in the 2000 households mixed residential/commercial
private, small and large companies. This can also lead to V2G and
suburb supplied by feeder #TDE05 [41], a static upper power limit
G2V strategies to assist the public grid when cars remain parked
is introduced. These electric vehicles will dispatch power to the grid
during working hours. We will show that with PV integration on
to avoid power demands surpassing this upper limit. Applying a
buildings, there is a large potential for solar electricity to supply all
merit-order strategy, the upper limit can be interpreted as the point
the energy needs of a large fleet of EVs.
where V2G-energy is cheaper than dispatching other reserves,
including transmission and distribution costs, in a strategy
2.5. Strategy 1: profit maximising with local electricity pricing designed to avoid overloads and promote grid stabilisation. In this
model case, grid stabilisation is a service provided by the car owner to the
utility, and the costs for this service are estimated. Discharge power
The Brazilian Electricity Regulatory Agency (ANEEL) released for each car is calculated with a modification of Eq. (2). Here Dt is
information about a new pricing model to be adopted starting in used, because there is no time limitation like in the first strategy:
2014 [38]. To reduce peak demand, a three-level tariff is going to be
established for residential users. The energy price during the peak    
interval from 18:30 to 21:30 is going to be 5 higher than the off- ESoc ðnÞ 1DODMax ECapa PMax
PDischarge ðnÞ ¼ min ; (9)
peak price as already mentioned. One hour before and 1 h after the Dt hConverter
peak interval, the energy price is going to be 3 higher than during
the off-peak period. Together with the recently approved net- As many cars as possible discharge with the highest discharge
metering regulation for PV generation on individual consumer power, which is limited by the grid connection, until the new de-
units of up to 1 MWp [24] in Brazil, these time-of-use tariffs can mand curve is below the upper limit, or there are no more cars
contribute to grid stabilisation when associated with an EV that is available. The battery charge level is updated according to Eq. (3).
The new demands are calculated as per Eq. (4).

2.7. Calculation of costs caused by battery degradation

Each day the costs for battery degradation by cycling are

calculated separately with and without discharge caused by the
V2G-service. At first the DOD is calculated

ESoC ðnÞ
DNew ðnÞ ¼ 1  (10)
Fig. 4. Efficiency and power flows.
L. Drude et al. / Renewable Energy 68 (2014) 443e451 447

Then the number of estimated cycle lifetime NCycles with the 8 GWh/year, enough electricity to supply over 4000 EV units, each
calculated DOD is calculated using Eq. (1). driving more than 12,000 km/year.
The costs for this discharge are then calculated with Braun and Rüther [42] have previously shown the role of grid-
connected, building-integrated PV generation in commercial
PriceBattery building energy and power loads in a warm and sunny climate. In
CostDegradation ðnÞ ¼ CostDegradation ðn  1Þ þ (11)
NCycles this work we show that not only can PV systems installed on res-
idential roofs contribute strategically to alleviate urban distribution
where PriceBattery is drawn from Table 1. The sums with and networks, shifting peak demand when there is a good correlation
without V2G-service are stored separately to be compared between loads and the solar radiation resource availability, but they
afterwards. can also supply all the energy requirements of a number of EVs that
is considerably larger than the number of residential households in
3. Results the corresponding suburb.

3.1. Contribution of photovoltaic generation 3.2. Strategy 1: profit maximising with local electricity pricing
In order to supply the new annual energy demands (MWh/
year) resulting from a proposed fleet of some 250 EVs in the The strategy of discharging cars only during peak hours in the
mixed residential/commercial suburb under study, the incorpo- early evening when tariffs are highest, leads to high revenues for
ration of rooftop PV generators on the single-family houses in car owners and avoided costs for the distribution utility. This leads
the suburb was proposed. Santos and Rüther [27] have identified to a new demand profile as shown in Fig. 6. Unfortunately, and not
the potential of the roofs in that area, which resulted in some surprisingly, two new peaks are resulting, right before and right
79,000 m2 of total roof surface, and where more than 50% of the after dispatch, using this V2G grid-support strategy. Fig. 7 shows
area are ideally orientated and tilted for receiving at least 95% of the simulated hourly demand profile of the urban feeder #TDE05
the theoretical maximum solar irradiation daily average density resulting from both the 7.9 MWp of PV generation (daytime hours
of 474 kWh/m2. Only areas where at least 1 kWp of PV could be feed-in), and EVs V2G grid support in the early evening (during the
integrated have been used in the simulation, resulting in more distribution utility peak hours) for a typical day in June 2009, as
than 43,000 m2, where 7.9 MWp of rooftop PV capacity can be well as the EVs recharging demands in the early hours (00:00e
incorporated. 06:00). It is clear that the EV contribution in the distribution util-
Fig. 1 has shown the high-energy consumption profile during ity’s official early evening peak hours can be anticipated to 18:00 to
the day (between 8:00 and 18:00), especially during the warmer make an optimised use of the EVs potential, for a larger number of
and sunnier summer months. The seasonality of demand shows a EVs might be available for V2G before 18:00. The incorporation of
good match with the photovoltaic generation profile; the months PV to generate the new energy demands represented by the new EV
with the largest energy demands are also the sunny summer fleet is comfortably accommodated during the daytime hours.
months. Fig. 5 shows that the integration of PV roofs on residential Figs. 6 and 7 also show that the early hours recharging of the EVs
buildings in the region supplied by feeder #TDE05 can drastically has to be closely monitored to avoid overloading the grid. The new
offset the energy consumption especially during summer months. energy demands are represented for a 250 EVs fleet, which is the
Fig. 5 also leads to the conclusion that during daytime hours this optimal value for cars participating in grid stabilisation as will be
feeder is able to accommodate these large amounts of photo- further shown.
generated electricity from grid-connected, rooftop PV generators The annual revenue as a function of DOD for an individual EV is
without the need of any additional distribution system upgrade. plotted in Fig. 8. With a DOD of 40% the revenue already exceeds
The potential of photovoltaic found for the region supplied by the US$ 1000 per year, where the annual revenue is calculated as the
#TDE05 feeder was 7.9 MWp, using high efficiency, commercially difference between the annual income and the annual costs. The
available crystalline silicon solar modules, and should be regarded income does not increase with a higher DOD after 75%, because of
as an upper limit for PV integration in that building stock. The the power limit PMax given by the grid connection and thus the
corresponding total annual PV generation expected is just under maximum DOD cannot even be reached.

24 10
24 10 22
22 20
20 18
18 16
Time (hours)

Power (MW)

16 14 6
Time (hours)

Power (MW)

14 6 12
12 10 4
10 4 8
8 6
6 4 2
4 2
2 0 0
0 50 100 150 200 250 300
50 100 150 200 250 300 Date (days)
Date (days)
Fig. 6. Strategy 1: simulated demand profile of the #TDE05 feeder in 2009 with the
Fig. 5. Simulated new demand profile of the #TDE05 feeder with the solar power new demands represented by the early hours (00:00e06:00) charging of 250 EVs, the
contribution of 7.9 MWp of PV generators spread over residential rooftops in the contribution of 7.9 MWp of PV generation during daytime hours, and the V2G
corresponding area. Data were calculated using the real demand profile for the year contribution of the same 250 EV fleet to grid support in the peak hours of the early
2009 and the corresponding solar irradiation profile for the same year. evening (19:00e21:00) at 75% maximum DOD.
448 L. Drude et al. / Renewable Energy 68 (2014) 443e451

12 3000
Demand with G2V and V2G
10 Demand with PV contribution
Original demand 2500

Accumulated Costs (US$)

Power (MW)

6 2000


0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hours)
500 Converter costs
Fig. 7. Strategy 1: simulation example of the new demand profile of the #TDE05 feeder Price of recharged energy
on June 12th, 2009 with the new demands represented by the early hours (00:00e Battery degradation costs
06:00) charging of the 250 EVs proposed fleet, the contribution of 7.9 MWp of PV 0
generation during daytime hours, and the V2G contribution of the same 250 EV fleet to 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
grid support in the peak hours of the early evening (19:00e21:00) at 75% maximum DOD
Fig. 9. Strategy 1: accumulated annual costs for V2G and G2V as a function of the DOD
for an individual EV.
The costs are the sum of the additional hardware needed for
V2G shown in Table 1 divided by 10 years, the price of energy
recharged for G2V, and the battery degradation caused by V2G, as value when compared with the 6949 residential and 1099 com-
shown in the stacked area plot in Fig. 9. The cost for the grid mercial consumer units in the area in 2010 [25].
connection does not increase with increasing the maximum DOD.
The money paid for recharged G2V energy rises linearly with MSE ¼ ðDðnÞ  Dav Þ2 ; N ¼ 30912 (12)
increasing the maximum DOD and reaches a maximum when the n¼0
power limit of the grid connection is reached. The battery degra-
dation increases nonlinearly with the DOD, as Fig. 2 implies. The MSE values are displayed as a colour map in Fig. 10, where
It turns out that the new pricing model proposed by the gov- the number of EVs used is varied along the horizontal axis, and the
ernment shows a large drawback. With an increasing market share maximum DOD is varied along the vertical axis. N ¼ 30,912, cor-
of storage using the contract strategy proposed in this work, grid responding to the 15 min time-steps over the 8760e1032
stability will suffer. To point out this effect, the corresponding mean (excluding the 43 missing days) hours in 2009. The MSE without
squared error was calculated according to Hartmann and Özdemir V2G-activity is calculated as 2.91 MW2. Values below this reference
[19], as shown in Eq. (12), where the upper limit N is the number of point represent a higher grid stability. It turns out that there is a
time-steps within a year. The mean square is a measure for the maximum grid stability of 2.57 MW2, reached with 250 cars and a
deviation between the average demand Dav and the current de- maximum DOD of 75%. Thus, optimising the car owners’ revenue
mand D(n). The optimal demand profile is a constant, thus no po- does not necessarily increase grid stability, and these conflicting
wer margins for higher demands have to be provided. The MSE is interests should be properly addressed by government-induced
zero for DðnÞ ¼ Dav ¼ const: energy policies regarding EVs and V2G strategies.
If the demand varies a lot, a very robust grid with high power In Fig. 6 the new demand profile as a result of Strategy 1 was
reserves has to be established yielding higher investment costs. If shown. Recharge and discharge intervals are clearly seen as hori-
there is a high variation in demand, the MSE value is very high as zontal stripes where the breaks are caused by time-changes
well. Thus a high MSE value is bad for grid stability. The error was because of daylight saving. The figure points out that even during
calculated for simulations with the number of EVs varying between times of low demand, coloured in blue, energy is delivered from
0 and 1000, and the DOD varying between 0.2 and 1.0. The EVs to the grid, which does not contribute to grid-stabilisation, but
maximum number of cars analysed is 1000, which is a rather high generates revenue for the EV owners.

1 5
Annual Revenue (US$)

3000 0.8
MSE (MW 2)


0 3

-1000 0.2
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 2.5
0 200 400 600 800 1000
DOD Cars

Fig. 8. Strategy 1: annual revenue for one EV as a function of the DOD with the Fig. 10. Strategy 1: mean squared error for different amounts of electric cars and
#TDE05 feeder in 2009. different DOD levels.
L. Drude et al. / Renewable Energy 68 (2014) 443e451 449

3.3. Strategy 2: grid stabilisation 12

Demand with G2V and V2G
In our Strategy 2 we aim at stabilising the distribution grid, 10 Demand with PV contribution
making more use of the V2G approach by applying a deeper DOD Original demand
level when necessary, while allowing the #TDE05 feeder to reach 8

Power (MW)
the maximum load limit of 6.25 MW before enabling V2G at the
utility’s peak hours time. Fig. 11 shows the resulting effect on the
feeder’s load profile in 2009. Fig. 12 shows the same typical day as
Fig. 7 (June 12th, 2009), using Strategy 2 instead. It shows that there
is a lower upper limit and there are no large peaks created by G2V, 2
as seen in Fig. 11. Analysing the demand data with Strategy 2, it
turns out that there are several days where the V2G-service is not 0
used at all, because the 6.25 MW utility demand limit for feeder 0 2 4 6 8 10 12 14 16 18 20 22 24
#TDE05 is not reached. There is also a small number of days when Time (hours)
the EVs’ battery capacity or discharge power limit are not large Fig. 12. Strategy 2: simulation example of the new demand profile of the #TDE05
enough to reduce the demand to the upper limit, resulting in a feeder on June 12th, 2009 with the new demands represented by the early hours
remaining peak. Nevertheless this strategy will work well, if the (00:00e06:00) charging of the 250 EVs proposed fleet, the contribution of 7.9 MWp of
dispatch of spinning-reserves is more expensive than the dispatch PV generation during daytime hours, and the V2G contribution of the same 250 EV
fleet to grid support in the peak hours of the early evening (19:00e21:00) at 75%
of V2G-services. The steep drop in the feeder #TDE05 demand
maximum DOD, only when the distribution utility’s 6.25 MW upper demand limit is
curve at noon shown in Fig. 12 indicates a large potential for EV reached.
recharging (G2V) at the workplace for the electric cars fleet.
In this case it also has to be noted that EV owners’ revenue is not
maximized, in contrast to Strategy 1. Again the overall grid stability reasonable compromise between costs and grid stability can be
is measured using Eq. (12) to compare Strategy 2 with Strategy 1. observed between 4.0 MW and 5.5 MW.
For 250 electric cars the minimum reachable MSE value is To compare both strategies, the MSE is plotted in Fig. 15 as a
2.57 MW2 for Strategy 1, and 2.55 MW2 for Strategy 2, with an function of the annual service cost per EV. It can be observed that
upper limit of 5 MW. Even if this difference is small, better using Strategy 2, a lower MSE, and thus a higher grid stability, can
knowledge of driving behaviour will lead to greater improvement be achieved. In addition, Strategy 2 fits well in the concept of merit
of MSE values for Strategy 2, but not for Strategy 1. order, whereas Strategy 1 does not. In Fig. 11 new energy demands
A comparison of different upper limits and their impact on grid caused by Strategy 2 are exposed. During times of low demands, no
stability evaluated with MSE is given in Fig. 13, where the upper energy is delivered to the grid. During times of high demands,
limit is varied between 4 MW and 8 MW. It turns out that the EVs peaks are shaved to the upper limit of 5e6 MW. In times of
contribute best to grid stabilisation in terms of MSE with an upper extraordinary high demands, battery capacity even at high DOD
limit of 5 MW. levels is not enough to shave peaks sufficiently, as seen during the
From the grid operator’s perspective, the upper limit should be days 61 and 62 at night in Fig 11.
chosen so that the dispatch of EVs is cheaper or equal to the
dispatch of other resources.
4. Energy balance
Calculating the service costs as the sum over the years of grid
connection hardware, costs for efficiency losses and battery
In this section an overall energy balance is discussed. Both
degradation divided by V2G-energy delivered, leads to Fig. 14. It can
Tables 2 and 3 present the energy balance along the year for
be observed that the service cost per delivered energy is very high
Strategy 1 and Strategy 2, respectively. Table 2 shows the energy
for a high upper limit, because the fixed costs have to be covered
balance considering Strategy 1, with 250 electric cars participating
without selling large amounts of energy to the grid. Furthermore
in grid stabilisation, with a 75% maximum DOD level, as well as the
the service cost for low upper limits is high, because of intensive
PV contribution of 7.9 MWp of rooftop generators spread all over
battery degradation. Comparing Figs. 13 and 14, it turns out that a
the metropolitan region supplied by the #TDE05 feeder. Photo-
voltaic energy is sufficient to compensate for energy losses due to
24 10
20 2.9
Time (hours)

Power (MW)

14 6 2.8

10 4
8 2.7
4 2
0 0
50 100 150 200 250 300
Date (days)
Fig. 11. Strategy 2: simulated demand profile of the #TDE05 feeder in 2009 with the 4 4.5 5 5.5 6 6.5 7 7.5 8
new demands represented by the early hours (00:00e06:00) charging of 250 EVs, the Upper Limit (MW)
contribution of 7.9 MWp of PV generation during daytime hours, and the V2G
contribution of the same 250 EV fleet to grid support in the peak hours of the early Fig. 13. Strategy 2: grid stability as a function of the upper limit (in MW), where V2G is
evening (19:00e21:00) at 75% maximum DOD, only when the distribution utility’s active with 250 electric cars participating in grid stabilization, and a 75% maximum
6.25 MW upper demand limit is reached. DOD level.
450 L. Drude et al. / Renewable Energy 68 (2014) 443e451

1.5 Table 2
Energy balance considering valid days in 2009 and Strategy 1.
Energy Service Costs (US$/kWh)

Months Valid V2G G2V PV Balance

days [MWh] [MWh] [MWh] [MWh]
January 29 95.7 158.6 799.7 736.9
February 28 92.4 157.1 689.0 624.4
March 29 95.7 162.7 838.5 771.6
April 29 95.7 162.7 748.6 681.6
0.5 May 31 102.3 173.9 682.8 611.1
June 18 59.4 101.0 411.3 369.8
July 16 52.8 89.7 293.8 256.8
August 24 79.2 134.6 609.7 554.3
September 27 89.1 151.4 491.3 429.0
4 4.5 5 5.5 6 6.5 7 7.5 October 30 99.0 168.3 664.6 595.3
Upper Limit (MW) November 30 99.0 168.3 797.1 727.9
December 31 102.3 173.9 837.2 765.6
Fig. 14. Strategy 2: service costs for V2G-service for each electric car per energy as a Annual 1062.6 1802.2 7863.6 7124.0
function of the upper limit where V2G is active, and a 75% maximum DOD level.
Balance: (V2G þ PV  G2V). Rows and columns may not add horizontally or verti-
cally due to rounding.
conversion and battery cycling efficiency, as well as the energy
demands caused by EV driving. As expected, PV energy generation
is higher in the summer season, and thus the balanced energy is 5. Conclusion
high during December and January. Over the year, Strategy 1 leads
to 1003 MWh of electricity supplied to the grid with V2G, We used real electricity demand data, and real local solar irra-
1801 MWh of electricity drawn from the grid with G2V, and diation and PV generation data to assess the potential contribution
7863 MWh of electricity generated by the 7.9 MWp of PV rooftops. of using both PV generation and V2G strategies to assist the dis-
There is therefore a considerably large positive annual balance of tribution utility #TDE05 feeder in peak shaving in the early eve-
electricity fed to the public grid (V2G þ PV generation  G2V) of ning. Residential rooftop PV generation spread over the area
7065 MWh. supplied by this utility feeder allowed to produce electricity well in
In Table 3 the energy balance for Strategy 2 with 250 cars excess of the additional amounts of electricity required by a num-
participating in grid stabilization, a 75% maximum DOD level, the ber of EVs to be owned by a fraction of the local residents.
PV contribution of 7.9 MWp of rooftop generators spread all over On the basis of a simulation with time-steps of 1 h using
the metropolitan region supplied by the #TDE05 feeder, and an measured demand data of the year 2009, PV systems have shown to
upper limit of 5 MW for the feeder demand is presented. It turns have the potential to reduce load fluctuations in the urban grid. The
out that the overall energy gain is larger than with Strategy 1. power generated correlates well with the demand data and might
Strategy 1 is optimised for electric car owners’ revenues, and the thus reduce the daytime demand peak. We pointed out that a
losses are thus not necessarily minimised. With Strategy 2 fewer system of 7.9 MWp installed in an urban region like the Santa
losses occur because the V2G option is used only to stabilise the Mônica suburb may compensate the consumption of air-
grid, and not to maximise car owners’ revenues. Over the year, conditioning in the commercial buildings.
Strategy 2 leads to 936 MWh of electricity supplied to the grid with This work points out that there is a potential for EVs to stabilise
V2G, 1747 MWh of electricity drawn from the grid with G2V, and the grid in regards of peak-load shaving. The new policy to be
the same 7863 MWh of electricity generated by the 7.9 MWp of PV implemented in Brazil in 2014, with higher energy prices at peak
rooftops. The positive annual balance of electricity fed to the public times also for residential consumers, should in fact result in
grid (V2G þ PV generation  G2V) is in this case very similar to the reduced demand during peak time, but might also create new
one resulting from Strategy 1, totalling 7053 MWh. In both cases, peaks around this time interval. As shown with Strategy 1, assumed
the surplus electricity generated by making use of the full PV po- there is a net-metering contract like the one recently established by
tential of the local residential rooftops is enough to feed an EV fleet the Brazilian regulatory agency for distributed micro (up to
of more than 4000 cars. 100 kW) and mini (up to 1 MW) power generating plants, this

Table 3
3 Energy balance considering valid days in 2009 and Strategy 2.
Strategy 1 Months Valid V2G G2V PV Balance
2.9 Strategy 2 days [MWh] [MWh] [MWh] [MWh]

January 29 64.3 119.6 799.7 744.4

MSE (MW 2)

February 28 88.0 151.7 689.0 625.3

March 29 97.7 165.1 838.5 771.1
April 29 91.4 157.3 748.6 682.7
2.7 May 31 102.3 173.8 682.8 611.3
June 18 59.0 100.4 411.3 369.9
July 16 53.9 91.1 293.8 256.6
2.6 August 24 80.8 136.7 609.7 553.8
September 27 89.0 151.3 491.3 429.0
October 30 20.4 165.6 664.6 519.4
0 200 400 600 800 1000 1200 November 30 96.9 168.7 797.1 725.3
Annual Service Cost Each Car (US$) December 31 93.0 166.0 837.2 764.2

Annual 936.7 1747.3 7863.6 7053.0

Fig. 15. Comparison between Strategy 1 and Strategy 2 regarding MSE, plotted as a
function of annual costs per electric car including all V2G hardware, battery degra- Balance: (V2G þ PV  G2V). Rows and columns may not add horizontally or verti-
dation and energy losses. cally due to rounding.
L. Drude et al. / Renewable Energy 68 (2014) 443e451 451

policy might be exploited leading to revenues for car owners, but [19] Hartmann N, Özdemir E. Impact of different utilization scenarios of electric
vehicles on the german grid in 2030. J Power Sources 2011;196:2311e8.
not necessarily to grid stabilisation. This effect is sensitive to the
[20] Gass V, Schmidt J, Schmid E. Analysis of alternative policy instruments to
number of EVs into consideration. promote electric vehicles in Austria. Renew Energy; 2012 (in press). Available
The second strategy, with a fixed limit, allows for the concept of online 6 September 2012.
merit order, and is therefore more suitable for the utility. To make [21] Kempton W, Tomic J, Letendre S, Brooks A, Lipman T. Vehicle-to-grid power:
battery, hybrid, and fuel cell vehicles as resources for distributed electric
use of this strategy, car owners would need to receive a small power in California. Institute of Transportation Studies, UC Davis; 2001.
reward just for having the car connected to the grid. Only in rare [22] Naspolini HF, Militão H, Rüther R. The role and benefits of solar water heating
cases, when demand exceeds a specific limit, the stored energy in the energy demands of low-income dwellings in Brazil. Energy Convers
Manage 2010;51:2835e45.
would be solicited by the utility. [23] Naspolini HF, Rüther R. The impacts of solar water heating in low-income
households on the distribution utility’s active, reactive and apparent power
Acknowledgements demands. Solar Energy 2011;85:2023e32.
[24] ANEEL e Agência Nacional de Energia Elétrica. Tarifa branca ao consumidor de
baixa tensão valerá com novo medidor. Available at:
The authors acknowledge the local distribution utility CELESC br/aplicacoes/noticias/Output_Noticias.cfm¼0pt; 2012;
for access to their feeder demand data. R. Rüther wishes to Agência Nacional de Energia Elétrica. Resolução Normativa 482/2012. Avail-
able at:; 2012.
acknowledge with thanks the Alexander von Humboldt Foundation [25] Fernandes M. Celesc e Centrais Elétricas de Santa Catarina S.A. Personal
for sponsoring the PV installation from which the PV generation communication; 2010.
performance data used in this work were obtained. Financial sup- [26] Dallinger D, Krampe D, Wietschel M. Vehicle-to-grid regulation based on a
dynamic simulation of mobility behavior. Working Papers Sustainability and
port from the Brazilian Research Council CNPq is also gratefully
Innovation S4/2010. Fraunhofer Institute for Systems and Innovation Research
acknowledged. (ISI); 2010.
[27] Santos IP, Rüther R. The potential of building-integrated (BIPV) and building-
References applied photovoltaics (BAPV) in single-family, urban residences at low lati-
tudes in Brazil. Energy Build 2012;50:290e7.
[28] Rüther R. Experiences and operational results of the first grid e connected,
[1] BEN e Balanço Energético Nacional. Available at:; building - integrated, thin film photovoltaic installation in brazil. In: Pro-
2012. ceedings of the second world conference and exhibition of photovoltaic solar
[2] Saitoh TS, Yamada N, Ando D, Kurata K. A grand design of future electric energy conversion, Vienna, Austria; 1998. pp. 2655e8.
vehicle to reduce urban warming and CO2 emissions in urban area. Renew [29] Rüther R, Dacoregio MM. Performance assessment of a 2 kWp grid-connected,
Energy 2005;30:1847e60. building-integrated, amorphous silicon photovoltaic installation in Brazil.
[3] Egbue O, Long S. Barriers to widespread adoption of electric vehicles: an Prog Photovolt Res Appl 2000;8:257e66.
analysis of consumer attitudes and perceptions. Energy Policy 2012;48:717e [30] Rüther R, Dacoregio MM, Salamoni IT, Knob PJ, Bussemas U. Performance of
29. the first grid-connected, BIPV installation in Brazil over eight years of
[4] Weiss M, Patel MK, Junginger M, Perujo A, Bonnel P, van Grootveld G. On the continuous operation. In: Proceedings of the 21st European photovoltaic solar
electrification of road transport e learning rated and price forecasts for energy conference, Dresden, Germany; 2006. pp. 119e22.
hybrid-electric and battery-electric vehicles. Energy Policy 2012;48:374e93. [31] CELESC. Centrais Elétricas de Santa Catarina S.A. Estado. Lei estadual 7.547, de
[5] Kempton W, Letendre SE. Electric vehicles as a new power source for electric 27/01/89; regulamentada pelo decreto 2.870, de 28/08/01, art. 26. Available
utilities. Transport Res D Transp Environ 1997;2:157e75. at:¼com_
[6] Kempton W, Kubo T. Electric-drive vehicles for peak power in Japan. Energy content&task¼view&id¼329&Itemid¼93; 2013.
Policy 2000;28:9e18. [32] Tomi c J, Kempton W. Using fleets of electric-drive vehicles for grid support.
[7] Kempton W, Tomi c J. Vehicle-to-grid power fundamentals: calculating ca- J Power Sources 2007;168:459e68.
pacity and net revenue. J Power Sources 2005;144:268e79. [33] Receita Federal. Simulador do tratamento tributário e administrativo das
[8] Yan X, Patterson D. Novel power management for high performance and cost importações. Available at:¼0pt; 2012.
reduction in an electric vehicle. Renew Energy 2001;22:177e83. [34] Biere D, Dallinger D, Wietschel M. Ökonomische analyse der erstnutzer von
[9] Ekman CK. On the synergy between large electric vehicle fleet and high wind elektrofahrzeugen. Z Energie 2009;33:173e81.
penetration e an analysis of the Danish case. Renew Energy 2011;36:546e53. [35] Rosenkranz C. Deep cycle batteries for plug-in hybrid application. In: EVS-20
[10] Ford A. Electric vehicles and the electric utility company. Energy Policy plug-in hybrid workshop, Long Island-CA, United States; 2003.
1994;22:555e70. [36] Meinhardt M. Pv-systemtechnik e ein motor der kostenreduktion für die
[11] Kempton W, Tomi c J. Vehicle-to-grid power implementation: from stabilising photovoltaische stromerzeugung. Niestetal: SMA Technologie AG; 2007.
the grid to supporting large-scale renewable energy. J Power Sources Available at:
2005;144:280e94. vortraege_jt2007/th2007_15_meinhardt.pdf.
[12] Sovacool BK, Hirsch RF. Beyond batteries: an examination of the benefits and [37] GIPA. Pesquisa de deslocamento de automóveis. Brazilian Interprofessional
barriers to plug-in hybrid electric vehicles (PHEVs) and a vehicle-to-grid Group in Automotive Products and Services; 2006. Technical report. Available
(V2G) transition. Energy Policy 2009;37:1095e103. at:
[13] Weiller C. Plug-in hybrid electric vehicle impacts on hourly electricity demand [38] Bittencourt R. Aneel aprova tarifas diferenciadas por horários de consumo.
in the United States. Energy Policy 2011;39:3766e78. Available at:
[14] Zhang Q, Tezuka T, Ishihara KN, Mclellan BC. Integration of PV power into diferenciadas-por-horarios-de-consumo; 2011.
future low-carbon electricity systems with EV and HP in Kansai Area, Japan. [39] Camargo I, Oliveira DM, Figueiredo F. Impact of the Brazilian renewable en-
Renew Energy 2012;44:99e108. ergy policies on the electrical energy crisisIn Power Engineering Society
[15] Mullan J, Harries D, Bräunl T, Whitely S. Modelling the impacts of electric Winter Meeting, vol. 1; 2002. pp. 393e4.
vehicle recharging on the Western Australian electricity supply system. En- [40] Losekann L, Oliveira A. Technology mix in the brazilian electricity sector. In:
ergy Policy 2011;39:4349e59. Proceedings of the 27th USAEE/IAEE North American Conference, Ann Arbor-
[16] Galus MD, Zima M, Andersson G. On integration of plug-in hybrid electric MI, Unites States; 2006.
vehicles into existing power system structures. Energy Policy 2010;38: [41] Pereira Junior LC. A interação entre geradores solares fotovoltaicos e veículos
6736e45. elétricos conectados à rede elétrica pública. Master’s thesis. Florianópolis-SC:
[17] Perujo A, Ciuffo B. The introduction of electric vehicles in the private fleet: Universidade Federal de Santa Catarina; 2011.
Potential impact on the electricity supply and on the environment. A case [42] Braun P, Rüther R. The role of grid-connected, building-integrated photovol-
study for the Province Milan, Italy. Energy Policy 2010;38:4549e61. taic generation in commercial building energy and power loads in a warm and
[18] Camus C, Farias T, Esteves J. Potential impacts assessment of plug-in electric sunny climate. Energy Convers Manage 2010;12:2457e66.
vehicles on the Portuguese energy market. Energy Policy 2011;39:5883e997.