Вы находитесь на странице: 1из 12

www.ietdl.

org
Published in IET Generation, Transmission & Distribution
Received on 5th September 2010
Revised on 27th February 2011
doi: 10.1049/iet-gtd.2010.0574

ISSN 1751-8687

Smart load management of plug-in electric vehicles in


distribution and residential networks with charging
stations for peak shaving and loss minimisation
considering voltage regulation
A.S. Masoum1 S. Deilami2 P.S. Moses2 M.A.S. Masoum2 A. Abu-Siada2
1
Operational Asset Management, Western Power, Kewdale, WA 6105, Australia
2
Department of Electrical and Computer Engineering, Curtin University of Technology, Perth, WA 6845, Australia
E-mail: paul.s.moses@gmail.com

Abstract: New smart load management (SLM) approach for the coordination of multiple plug-in electric vehicle (PEV) chargers
in distribution feeders is proposed. PEVs are growing in popularity as a low emission and efficient mode of transport against
petroleum-based vehicles. PEV chargers represent sizeable and unpredictable loads, which can detrimentally impact the
performance of distribution grids. Utilities are concerned about the potential overloads, stresses, voltage deviations and power
losses that may occur in distribution systems from domestic PEV charging activity as well as from newly emerging charging
stations. Therefore this study proposes a new SLM control strategy for coordinating PEV charging based on peak demand
shaving, improving voltage profile and minimising power losses. Furthermore, the developed SLM approach takes into
consideration the PEV owner preferred charging time zones based on a priority selection scheme. The impact of PEV
charging stations and typical daily residential loading patterns are also considered. Simulation results are presented to
demonstrate the significant performance improvement offered by SLM for a 1200 node test system topology consisting of
several low-voltage residential networks populated with PEVs.

1 Introduction home from work within a narrow time period and


immediately plug-in their vehicles to charge during a time
Plug-in electric vehicles (PEVs) are growing in popularity in of already high-peak demand. PEV charging is also
developing nations in an effort to overcome the problems of expected to take place in public or corporate car parks, and
pollution, depleting natural oil and fossil fuel reserves and electric vehicle charging stations, which employ high-power
rising petrol costs. Furthermore, automotive industries are rapid charging facilities [6, 7]. These high consumption
facing increasing community pressure and governmental uncoordinated and random charging activities could
regulations to reduce emissions and adopt cleaner, more significantly stress the distribution system causing severe
sustainable technologies such as PEVs. Therefore several voltage fluctuations, suboptimal generation dispatch,
car manufacturers have already begun to expand into the degraded system efficiency and economy as well as
PEV market [1 – 4]. increasing the likelihood of blackouts because of network
In order to prepare for the large proliferation of PEVs into overloads. Therefore there is a recognised need for
modern power systems, distribution grids must evolve with monitoring and control of the network under such
new infrastructure to support an electric vehicle network. circumstances with modern Distribution Management
This is one of the initiatives brought about by newly Systems (DMS) at the primary substation levels [8 – 11].
developing smart grids. Through a sophisticated two-way Fortunately, the development of smart grids will provide
communications infrastructure with smart meters and an excellent opportunity to enhance DMS operations with
sensors, smart grids will enable real-time monitoring and intelligent or smart coordinated charging of PEVs. The
control of transmission, distribution and end-user consumer smart grid communication network will enable the DMS to
assets (e.g. smart appliances) for effective coordination and directly interact with many of these consumer loads (e.g.
usage of available energy resources. PEV chargers and smart appliances). In support of the
The operation of PEVs in a distribution system will be a DMS, a novel smart load management (SLM) approach
challenging demand side management (DSM [5]) problem is proposed for the control and coordination of multiple
from the utilities perspectives since PEV battery chargers PEV charging activity while reducing system stresses that
represent sizeable and difficult to predict loads. A quite can detrimentally impact grid reliability, security and
plausible scenario is that numerous PEV owners will arrive performance [12].

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877– 888 877
doi: 10.1049/iet-gtd.2010.0574 & The Institution of Engineering and Technology 2011
www.ietdl.org
SLM is proposed to coordinate PEV charger scheduling to on cable and transformer losses. Peak shaving is achieved by
perform peak demand shaving, improve voltage profile and minimising system demand using the following objective
minimise power losses. The SLM also takes into function
consideration the PEV owner’s preferred charging time zones
based on a priority selection scheme. To demonstrate the 
hend 
hend 
n
improvement in distribution system performance with SLM, min Demandh = min h
Pk,load (3a)
a detailed test system topology (with 1200 nodes) is h=hstart h=hstart k=1
simulated, which consists of a high-voltage 23 kV feeder (the
IEEE 31 bus test system) with several integrated low-voltage where hstart and hend correspond to the starting and ending
53 node 415 V residential networks populated with PEVs. charging hours within the selected charging time zone,
Simulation results including total power consumption, system h
respectively. Pk,load is the load demand of node k at hour h.
losses and voltage profiles over a 24 h period are presented The loss minimisation objective is defined as the
for (un)coordinated charging with PEV penetrations of 17, minimisation of incremental system losses within the 24 h
31, 46 and 62% considering three tiered charging time
zones: red, 1800–2200 h; blue, 1800–0100 h and green,

24
1800–0800 h. min Wloss = h
Ploss (3b)
h=1
2 Problem formulation: coordination of PEV
charging The power losses of the distribution system are computed
from the Newton-based power flow outputs. The power loss
The coordination problem of PEVs is to define the scheduling in each line section between nodes k and k + 1 is
of individual PEV chargers in a distribution system such that
total system power losses are minimised, peak demand
shaving is achieved and voltages at all nodes are regulated Ploss(k,k+1) = Rk,k+1 (|Vk,k+1 − Vk ||yk,k+1 |)2 (4)
within allowable tolerances. Existing residential load
variations over a 24 h period, grid topology, the effects of and the total power loss is
charging stations and PEV owner selected charging priorities
will also be considered in the PEV coordination approach. 
n−1
This section describes how the charging coordination Ploss = Ploss(k,k+1) (5)
problem can be formulated into two objective functions k=0
taking into account the necessary constraints to improve the
distribution system performance. 2.3 Objective power flow method and load model

2.1 Distribution system constraints A modified Newton-based load flow routine is implemented
to assess the state of the distribution system subject to PEV
The voltage constraints of the distribution system will be charging considering voltage profile and power losses,
considered by setting the upper and lower limits to which is necessary for the computation of the objective
correspond with voltage regulation limits typically set by function and checking of constraints. All loads are modeled
utilities. In this paper, the voltage limits are set to +10% as constant power loads with their real and reactive powers
(V min ¼ 0.9 pu and V max ¼ 1.1 pu), which is typical of updated through a load curve for each hourly time interval
many distribution systems the load flow is computed.

V min ≤ Vk ≤ V max , for k = 1, . . . , n (1) 3 SLM for the PEV coordination problem

where k is the node number and n is the total number of nodes. A new SLM approach for the coordination of PEV charging is
The second constraint is for limiting the total maximum system proposed. SLM is designed to operate as part of the DMS for
demand of the distribution system to prevent the occurrence of a distribution feeder to perform intelligent scheduling of PEV
an overload condition from PEV  charging. Therefore the total battery chargers considering distribution system performance.
power consumption at each hour ( Demandh) is limited to the This approach also considers human inputs by allowing the
peak demand level (Dmax) that would normally occur without priority of PEV charging to be selected by the PEV owner.
any PEVs connected in the residential networks This section describes the development of the overall SLM
methodology based on the objective function and system
 
n constraints defined in Section 2.
Demandh = h
Pk,load ≤ Dmax , for h = 1, . . . , 24 As an alternative to PEV chargers randomly and
k=1 immediately operating when first plugged in, or after some
(2) fixed time delay, the proposed SLM will decide which
PEVs will commence charging and at what time. PEV
2.2 Objective function for peak shaving and loss charger control can be achieved through the forthcoming
minimisation smart grid communications infrastructure by sending and
receiving signals to individual PEV chargers. This means
The selected objective functions for the PEV-charging that PEV charging control would be taken out of the hands
coordination problem is based on the minimisation of system of the owner and scheduled automatically. The proposed
demand and total system power losses over the charging SLM will perform peak shaving (3a), loss minimisation
hours. The justification for using losses as part of the (3b) and voltage regulation based on the system constraints
objective function is that a distribution systems economy will ([see (1) and (2)]. Furthermore, SLM also takes into
largely depend on the cost of energy that would be expended consideration existing load variations over a 24 h cycle,

878 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877 –888
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2010.0574
www.ietdl.org
The search process is started by running the load flow loop
systematically for individual PEV chargers temporarily
activated one at a time to find the PEV within that priority
group that will result in minimum system losses. For each
Fig. 1 Subscription options of charging time zones for PEV temporarily activated PEV charger in the search phase, the
owners sum of the losses throughout the entire charge duration is
considered and compared for all temporary PEV node
while factoring PEV owner preferences for charging time placements. If at any time the load flow indicates a
zone and priority. constraint violation (e.g. voltage out of limits) at any node
[see (1) and (2)], the algorithm will attempt to reschedule
3.1 Charging time zone and priority scheme the PEV and perform another scanning of the system load
curve until the constraints are satisfied. Therefore it may not
The developed SLM allows PEV owners to indicate their be possible for all PEV owners to be accommodated in
preferred charging time zone. The SLM will try to their preferred charging zones and must be deferred to the
accommodate these preferences in the charger scheduling, next possible hour.
while considering the peak shaving and loss minimisation Once it has been determined which PEV node in that
objective functions as well as system constraints (1 – 5). priority can begin charging at the selected time with
Three charging zones (Fig. 1) have been defined for this minimum system losses, the selected PEVs scheduling is
study: permanently assigned and the system load curve updated
ready for the next iteration. This process is repeated for all
1. Red charging zone (1800– 2200 h) is for (high priority) nodes in that priority group before advancing to the next
PEV owners wanting to charge their PEVs as soon as time zone (e.g. blue zone). At the end of this process, the
possible upon return from work to have their vehicles ready SLM algorithm arrives at individual schedules assigned to
for use later in the evening. Red charging zone coincides all PEV chargers. The programme then exits the main loop
with most of the on-peak period. Therefore PEV owners and computes the 24 h load flow to print new system
desiring to charge during this period of high demand will performances (e.g. all node voltage profiles and power
be charged a higher tariff rate. losses).
2. Blue charging zone (1800 – 0100 h) is for (medium
priority) consumers that prefer to charge their vehicles at
partially off-peak periods and pay a lower tariff rate.
3. Green charging zone (1800– 0800 h) is the period that
most PEV charging is expected to take place because most
(low priority) consumers will require their vehicles fully
charged for use throughout the next day. Charging off-peak
will be highly encouraged by setting the cheapest tariff rates.

Note that all charging zone periods begin at 1800 h and are
allowed to overlap for some duration. This makes it possible
for SLM to accommodate fortunate medium and low priority
subscribers the opportunity to charge their vehicles earlier
if possible. That is, after scheduling higher priority
subscribers, lower priority customers can be served if there
is enough capacity without violating system constraints.

3.2 Proposed SLM algorithm

A MATLAB-based algorithm has been developed to perform


PEV scheduling based on SLM (Fig. 2). After reading the
input parameters (e.g. node and branch impedance data,
specified nodes with PEVs, designated priority time zones,
load profiles for PEV chargers and residential loads as well
as system constraints) and initialisation (e.g. selecting the
highest priority group, time zone and PEV), the main
programmme loop begins with the highest priority group of
PEVs (e.g. red zone) and determines a suitable PEV
charger start time within that time zone. This is done by
scanning a predetermined system load curve (which is
updated at each iteration) to find the ideal time within the
priority group time zone corresponding to the least demand
[see (3a)]. Once an ideal start time has been selected, there
is still the question of which PEV node should begin
charging. This is an important consideration because the
node at which a PEV is charging will impact the load flow,
power losses and system voltage profile. Therefore the SLM Fig. 2 Proposed SLM algorithm for coordinated PEV scheduling
will search for the PEV node that will result in the least considering peak load shaving, loss minimisation, voltage
system losses [see (3b)]. regulation and peak demand limits

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877– 888 879
doi: 10.1049/iet-gtd.2010.0574 & The Institution of Engineering and Technology 2011
www.ietdl.org
4 Distribution system with PEVs and reasonable charging profiles necessary for the study. PEV
charging stations battery capacities typically range from a few kWhs to over
50 kWh [1 – 4, 14]. For this study, most of the simulations
A detailed smart grid system test topology is developed and assume a 10 kWh battery capacity per PEV. Selected
studied to compare the impacts and benefits of the proposed simulation cases are also performed with battery sizes
PEV charging approach (SLM) against random uncoordinated ranging from 5 kWh to 15 kWh. It is expected that the
charging. This section describes the system configuration and lower end of battery sizes are more affordable and more
load assumptions necessary for the analysis. likely to initially dominate the market.
In order to maximise PEV battery life and performance to
4.1 System topology achieve maximum number of charging cycles, deep cycle
batteries have a rated depth of discharge (DOD) which
The selected system is a hybrid of the IEEE 31 bus 23 kV should not be exceeded. In lithium-ion batteries currently
distribution test system [13] and several densely populated used in some PEVs, optimum DOD typically ranges from
residential LV 53 node 415 V feeders based on real system 50 to 80%, which is a tradeoff between battery life and
data of a neighbourhood (in Western Australia). The HV driving range. Therefore this study assumes a typical DOD
feeder has six branches extending into different districts with of 70% of the rated battery size resulting in an available
PEV charging stations also placed at selected HV nodes. LV capacity of 7 kWh that the charger must deliver in order to
sections each consist of 53 nodes representing customer charge a 10 kWh PEV battery. Battery chargers have some
households with selected nodes assigned PEVs, priority and losses and therefore the energy requirement from the grid is
charging zone (Fig. 3). A total of 22 LV 53 node residential actually greater than the stated battery capacity. A typical
feeders are implemented and are supplied from the HV main battery charger efficiency of 88% is assumed [15] thereby
buses via 23 kV/415 V 300 kVA distribution transformers. requiring a total of 8 kWh of energy from the grid in order
The total number of nodes of this system is 1200. The to charge a single PEV with a 10 kWh battery.
system impedances are listed in the Appendix.
4.3 Residential PEV battery chargers
4.2 PEV energy requirements
In practice, PEV battery chargers will have large power ratings in
For realistic modeling of PEV charging loads, the battery comparison to normal residential loads in order to charge the
capacities must be considered in order to determine large PEV battery banks within reasonable time periods.

Fig. 3 Designated PEV penetration levels and assigned priorities for charging time zones (red ¼ high priority, blue ¼ medium priority,
green ¼ low priority)
Boxes with no colour indicates nodes with no PEVs present

880 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877 –888
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2010.0574
www.ietdl.org
stations, will need to be deployed in various locations
around the network. PEV requirements from charging
stations are such that it would be necessary to draw
sufficient amount of power to charge PEV batteries in a
similar time duration to that of filling the tank of a
conventional fuel based vehicle. That is, PEVs pulling into
a charging station would need a large amount of energy
delivered in a short period of time for rapid recharging of
batteries which serve as their ‘fuel tank’ [7]. Therefore
large power rectification conversion equipment capable of
transforming high-power ac to high-power dc will be
necessary to deliver this amount of power.
In this analysis, the impact of rapid PEV charging stations
is considered at four specific sites near residential areas
populated with PEVs. The charging stations are connected
to HV main buses 2, 4, 7 and 9. These charging stations
follow a load curve as shown in Fig. 4. There are two
charging station load peaks; morning and evening which
Fig. 4 Daily load curves for residential loads and charging stations correspond to PEV owners charging their vehicles before
they go to work in the morning, and, after returning from
However, limitations of household wiring must also be work in the evening. The maximum peak demand of PEV
considered. A standard single-phase 240 V outlet (Australia) charging stations corresponds to the maximum number of
can typically supply a maximum of 2.4 kW. There are also 15 PEVs that can pull into a charging station during its busiest
and 20 A outlets (single-phase and three-phase) that can supply time, which is assumed to be on average six PEVs. The
approximately 4 and 14.4 kW, respectively. For this analysis, a PEV charging station power requirement for the rapid
maximum charging power level of 4 kW at unity power factor charging of one PEV is assumed to be approximately
is selected because this is commonly available in most single- 14.4 kW at unity power factor. Therefore the peak demand
phase residential households without having to reinforce wiring. of one charging station is approximately 86.4 kW.

4.4 PEV charging stations


4.5 Residential load profiles
In order for PEVs to be a viable option and achieve long
distance traveling similar to conventional fuel-based cars, Based on actual recordings from a distribution transformer
rapid charge electric vehicle service stations, or charging (in Western Australia), a typical residential load curve is

Fig. 5 1200 node smart grid distribution system topology consisting of the IEEE 31 bus 23 kV system with several 415 V residential feeders
Each low-voltage residential network has 53 nodes representing customer households with varying penetrations of PEVs (Fig. 3)

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877– 888 881
doi: 10.1049/iet-gtd.2010.0574 & The Institution of Engineering and Technology 2011
www.ietdl.org
constructed to model the domestic load variations (without
PEV charging) at each house over a 24 h period (Fig. 4).
The peak power consumption of a house is assumed to be
on average 1.5 kW with a power factor of 0.95 and occurs
in the evening at around 1800 h.

4.6 PEV penetration levels

In order to cover the widest range of plausible PEV charging


scenarios in the near and long term future, PEV penetration
levels of 17, 31, 46 and 62% are simulated for each
charging approach. The penetration levels are defined to be
the proportion of nodes with PEVs to the total number of
LV residential nodes (excluding transformer LV nodes). A
maximum of one PEV per household is assumed.

4.7 Designated PEV priorities

Within the different PEV penetrations assumed, PEVs are


grouped into three different priority time zones (e.g. red,
blue and green zones). A realistic breakdown of priorities is
assumed by having the majority of the PEV owners
subscribing to green and blue zone priorities. This is
because the lower pricing of blue and green zones will be
more attractive to PEV owners compared with higher tariffs
in red zone charging. The priority groups are randomly
allocated to PEV nodes in each LV residential network as
shown in Fig. 3.

5 Results and discussion


Simulation results based on the SLM approach described in
Section 3 (Fig. 2) and the 1200 node distribution system
topology of Section 4 (Fig. 5) are presented and discussed
in this section. Different charging scenarios for residential
networks populated with PEVs as well as the effects of four
charging stations located along the distribution feeder are
compared in Figs. 6 – 13 and Tables 1 and 2. Furthermore,
the impacts of random uncoordinated PEV charging against
coordinated charging based on the proposed SLM are also
studied.

5.1 Impact of random uncoordinated PEV charging

Figs. 6 – 8 show the results of three different random


uncoordinated PEV charging scenarios and the resulting
performance on the distribution system. PEVs are scheduled
randomly following a normal distribution over three different
charging time zones. As illustrated in Fig. 6, an optimistic
but unlikely scenario of PEV charging is shown where the
scheduling is spread over the widest time duration (1800–
0800 h). Despite keeping the system peak from increasing
drastically, voltage deviations beyond the regulation limits
occur at peak time with significant increases in system losses Fig. 6 Random uncoordinated PEV charging within the green
(Figs. 6b and c). This PEV charging scenario is unlikely to charging zone (1800–0800 h)
happen because most PEV owners will start plugging in their
a Total system power demand for 46% PEV penetration
vehicles soon after returning home from work until later in b Voltage profiles (shown for worst affected nodes) for different PEV
the night, with very few PEV charging occurring in the penetrations. Note that the high penetrations (e.g. .31%) cause moderate
morning hours. voltages deviations
The situation becomes worse when a more realistic c Total system power losses for different PEV penetrations
scenario is considered by performing uncoordinated PEV
charging over narrower time zones with PEVs charging
immediately on arrival from work, or, after some fixed over 1800 – 0100 h and 1800 – 2200 h, respectively. The
delay into the evening or later at night. The impact on system peak rises sharply and broadens because of much
distribution system performance is shown in Figs. 7 and 8 of the PEV charging load coinciding with normal
that depict random uncoordinated charging distributed residential load peaks. This can have a detrimental

882 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877 –888
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2010.0574
www.ietdl.org

Fig. 7 Random uncoordinated PEV charging within the blue Fig. 8 Random uncoordinated PEV charging within the red
charging zone (1800–0100 h) charging zone (1800–2200 h)
a Total system power demand for 46% PEV penetration a Total system power demand for 46% PEV penetration
b Voltage profiles (shown for worst affected nodes) for different PEV b Voltage profiles (shown for worst affected nodes) for different PEV
penetrations. Note that the high penetrations (e.g. .31%) cause moderate penetrations. Note that the high penetrations (e.g. .31%) cause moderate
voltages deviations voltages deviations
c Total system power losses for different PEV penetrations c Total system power losses for different PEV penetrations

impact on generation dispatching with limited spinning excessive power losses as well as significant increase
reserve to service this new load peak because of PEVs. in transformer loading (Table 1, Figs. 7b and c and 8b
Severe voltage deviations up to 20% are observed with and c).

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877– 888 883
doi: 10.1049/iet-gtd.2010.0574 & The Institution of Engineering and Technology 2011
www.ietdl.org

Fig. 10 Impact on system power demand with 62% penetration of


PEVs using SLM-coordinated PEV charging for a PEV battery size
of 10 kWh (Table 2)

5.2 Impact of SLM-coordinated PEV charging

The proposed SLM algorithm is applied in an attempt to


overcome the significant performance degradations observed
with uncoordinated random PEV charging. The results for
SLM coordinated PEV charging are demonstrated in
Figs. 9–13, and Tables 1 and 2. Coordinated PEV charging
without (Table 1) and with (Table 2) PEV owner preferred
time zone priorities is also considered. SLM results indicate
a significant improvement in distribution system performance.
Most notably, the system peak demand has been
significantly curtailed which is more favourable from a
standpoint of generation dispatch and preventing overloads
(e.g. Fig. 9a for 46% PEV penetration). This also has a
significant impact on the system efficiency by reducing I 2R
losses in cables and transformer impedances. The energy
losses to energy consumption ratio reduced to 2.6% (62%
PEV penetration) against the worst uncoordinated case with
3.37% losses. Furthermore, peak power losses are limited to
less than a third of the worst case random uncoordinated
charging as shown in Fig. 9c.

Fig. 9 Proposed SLM coordinated PEV charging considering


PEV owner priority charging zones for a PEV battery size of
10 kWh (Table 2)
Note that SLM schedules PEVs in the owner designated time zones, performs
peak load shaving and minimises system losses for different PEV penetrations
a Total system power demand for 46% PEV penetration
b Voltage profiles (shown for worst affected nodes) for different PEV
penetrations. Note that SLM maintains the voltages at all nodes within
regulation
c Total system power losses for different PEV penetrations. Note the
Fig. 11 Impact on system power demand with 31% penetration of
significant reduction in losses compared with the uncoordinated random PEVs using SLM- coordinated PEV charging for a PEV battery size
charging of 10 kWh (Table 2)

884 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877 –888
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2010.0574
www.ietdl.org
The SLM coordinated PEV charging also has positive
impacts on peak transformer load currents. This is an
important consideration as distribution transformers will be
the weakest link in residential networks and newly developing
smart grids populated with PEVs. For many of the
uncoordinated random charging scenarios (Table 1),
distribution transformers are experiencing significant increases
(e.g. over 200%) in peak loading which can affect reliability
and expected service life. With SLM coordination,
transformer currents are significantly reduced by at least 40%
when compared with the worst case uncoordinated charging.

5.3 Impact of PEV battery sizes on SLM


coordinated PEV charging

In the previous simulations, a fixed PEV battery size of


10 kWh was assumed. In this section, the proposed SLM
coordination has also been performed for cases with
Fig. 12 System power demand with 17% penetration of PEVs different PEV battery size assumptions ranging from 5 to
using SLM- coordinated PEV charging for a PEV battery size of 15 kWh capacities using the same charging efficiencies and
10 kWh (Table 2) depth of discharge. Therefore SLM must contend with the
various PEV battery sizes dictating different charger power
levels (e.g. 2.5– 4 kW) and charging durations (e.g., 2 – 3 h).
As shown in Table 2, similar performance improvements
can be achieved even with 50% additional PEV charging
loads (e.g. charging 15 kWh batteries) in residential
networks. The tradeoff in this case is that more PEVs in
their preferred priority charging zones may need to delay
their charging because the system is under more strain
leading to demand limitations and voltage constraints.

5.4 Further issues


The presented case studies simulate a wide range of scenarios;
however, there are still large uncertainties as to how PEV
charging will unfold in future smart grid distribution systems.
Specifically, future PEV loading patterns may not conform to
standard statistical distribution load models (e.g. Poisson or
Gaussian). For instance, the random nature of arrivals and
departures of PEVs in residential networks and their charging
Fig. 13 Special case with 38% penetration of PEVs where some activities will vary considerably along different buses in a
blue PEV owners can begin charging in the red zone because of distribution system. Some PEV owners may require fast
available capacity charging during the peak rush hours in spite of higher-energy
prices, whereas others might prefer to wait and charge their
SLM coordinated PEV charging is demonstrated to be vehicles at early morning and off-peak hours. To that end,
effectively regulating voltages at all nodes to comply with there is currently significant research being performed on
set regulation limits. The worst case unacceptable voltage developing accurate load distribution models using a variety
deviations of 19.38% at the worst node (Table 1) for of techniques such as the Gaussian mixture model (GMM)
uncoordinated PEV charging are compensated to less than proposed in [9–11]. The proposed SLM algorithms are not
10%. However, there is a tradeoff in that some PEV owners restricted to any particular load distribution and will be
who subscribed to a preferred priority charging time zone capable of coordinating PEV charging for a wide range of
were not accommodated in their requested charging time loading scenarios with little or no modifications necessary.
zone because the system reached a point where PEV Furthermore, SLM algorithms can be applied to the
loading caused voltage constraints to be violated. This may coordination of other deferrable loads and smart appliances
be the case if high priority PEV owners are charging in (e.g. smart appliances such as dishwashers, clothes dryers and
distant locations which are more sensitive to voltage drops washing machines).
and line losses. SLM handled these cases by attempting to Another challenge in the implementation of SLM in the
reschedule the PEV owners causing the violations to the next DMS is obtaining the necessary inputs (e.g. system losses,
available hour where the system was under less strain, thereby power flows and voltage profiles) to perform the PEV
satisfying constraints. This is evident for higher penetration coordination. In the proposed SLM, this information is
cases as shown in Figs. 9a and 10. On the other hand, if there obtained from load flow computations and simulated load
is available system capacity with minimal or no constraint profiles. Unlike transmission systems, the amount of real-
violations, SLM will allow simultaneous charging of lower time network measurements in distribution networks,
priority PEVs in a higher priority zone. A special case of this especially down to the low- voltage customer level, is
is shown in Fig. 13 where some blue zone subscribers are currently very limited. Smart grids propose to improve this
allowed to charge their vehicles in the red zone at 2100 h. situation by deploying smart meters and sensors in

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877– 888 885
doi: 10.1049/iet-gtd.2010.0574 & The Institution of Engineering and Technology 2011
www.ietdl.org
Table 1 Impact of uncoordinated and coordinated PEV charging on the power quality of the 1200 node distribution system (Fig. 5)

PEV penetration, % Uncoordinated charging (random charging) Coordinated charging based on peak shaving
without priority [see (1) –(3)]

D lossa, % D V b, % I MAX
c
, pu D lossa, % D V b, % I MAX c, pu

Case 1: No priority, charging period: 6 pm–8 am (Fig. 6)


17 2.2878 11.2646 0.32271 2.2214 7.9479 0.29488
31 2.4447 14.6738 0.35889 2.3286 9.6372 0.29488
46 2.538 14.213 0.36463 2.4113 9.8457 0.29488
62 2.6705 14.8894 0.41401 2.4952 9.9988 0.29488
Case 2: No priority, charging period: 6 pm–1 am (Fig. 7)
17 2.3711 11.6479 0.36167 2.3248 9.9956 0.30479
31 2.6135 13.9612 0.41267 2.53 9.9938 0.33868
46 2.8015 15.4523 0.46203 2.6275 9.9855 0.37814
62 3.0179 15.7275 0.52543 2.7203 9.9892 0.39967
Case 3: No priority, charging period: 6 pm–10 pm (Fig. 8)
17 2.4267 12.439 0.39635 2.3661 9.7661 0.37082
31 2.7469 15.6393 0.4869 2.5392 9.9915 0.38401
46 3.0359 17.6879 0.56379 2.6321 9.9917 0.40024
62 3.3714 19.384 0.66702 2.7249 9.9895 0.40851

a
Ratio of system losses over 24 h compared with total power consumption over 24 h
b
Voltage deviation at the worst node
c
Maximum of all distribution transformer load currents

Table 2 Impact of coordinated PEV charging based on SLM on residential networks that can telemeter data to the utility via a
the power quality of the 1200 node distribution system (Fig. 5) bidirectional communications network. This will significantly
enhance the ability of the DMS to monitor and control
PEV penetration, % Coordinated charging based on distribution system assets and consumer loads. Integral to
peak shaving with priority ((1)–(3), the processing of metered real-time network data will be the
Figs. 9–13) use of state estimation (SE) techniques.
Transmission systems already employ some level of SE
D lossa, % D V b, % I cMAX, pu with SCADA systems to gather information. However, the
5 kWh PEV battery size (2 h charging) operational philosophies and network topologies of
17 2.2151 8.7799 0.29488 distribution feeders vary considerably compared to
31 2.2506 8.2493 0.29488 transmission system operations. This is even more of an
46 2.2689 7.9479 0.30216 issue with the rollout of smart grids which represent a
62 2.3174 8.0024 0.30933 fundamental change to distribution system design. For
7.5 kWh PEV battery size (2 h charging) example, smart grids DMS must also contend with
17 2.2564 9.9369 0.29946 dynamic operation from increased renewable energy
31 2.3351 9.8199 0.29823 penetration, distributed generation activity as well as PEV
46 2.3680 9.2452 0.33045 charging at residential premises and large-scale charging
62 2.4493 9.9954 0.32905 stations. Therefore new distribution system state estimation
10 kWh PEV battery size (2 h charging) (Figs. 9–13) (DSSE) methods, which will be the heart of the DMS, are
17 2.3005 9.9943 0.30725 currently being explored [8, 10] as well as smart meters
31 2.4216 9.7064 0.30883 and sensors to enhance the observability and monitoring
46 2.4725 9.8929 0.33404 of the distribution system. For the SLM coordination
62 2.5967 9.9926 0.40727 problem, this means that the DSSE would serve in
12.5 kWh PEV battery size (3 h chargingd) place of the load flow step in the algorithm of Fig. 2 in
17 2.3132 9.9848 0.30200 order to supply the necessary data and feedbacks to
31 2.4295 9.9673 0.30155 coordinate PEVs and maximise feeder performance.
46 2.4947 9.5322 0.33606 Specifically, DSSE will be required to provide up-to-
62 2.6388 9.8761 0.33859 date voltage magnitudes at all nodes (e.g. for checking
15 kWh PEV battery size (3 h chargingd) voltage regulation), estimate total load consumption (e.g.
17 2.3560 9.9943 0.30725 to check demand limitations), sense and process individual
31 2.5205 9.9655 0.30883 PEV charger power profiles, and project system losses
46 2.6046 9.9972 0.33404 necessary for the SLM loss minimisation objective
62 2.7904 9.9546 0.36402 function. The above-mentioned complicated issues related
to PEV load modelling and DSSE are beyond the scope of
a
Ratio of system losses over 24 h compared with total power this paper and should be tackled in future publications.
consumption over 24 h
b

c
Voltage deviation at the worst node 6 Conclusions
Maximum of all distribution transformer load currents
d
Charging duration is increased in order to charge the larger This paper studies the impacts of various random
batteries uncoordinated charging scenarios as well as coordinated

886 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877 –888
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2010.0574
www.ietdl.org
PEV charging for a distribution system with charging stations 3 Chevrolet: ‘2011 volt electric car’. 2010. Available at: http://www.
and residential PEV charging activity. A novel SLM algorithm chevrolet.com/electriccar/
4 Mitsubishi: ‘i MiEV Mitsubishi innovative electric vehicle’. 2010.
is developed in Matlab environment for coordinating the Available at: http://www.mitsubishi-motors.com/special/ev/
scheduling of multiple PEVs while considering distribution 5 Gönen, T.: ‘Electric power distribution system engineering’ (Taylor &
and residential grid performances (e.g. voltage profile, Francis, Boca Raton, 2007)
system losses and peak demand shaving). Three charging 6 Clement-Nyns, K., Haesen, E., Driesen, J.: ‘The impact of charging
plug-in hybrid electric vehicles on a residential distribution grid’,
time zones (red, 1800– 2200 h; blue, 1800–0100 h and IEEE Trans. Power Syst., 2010, 25, (1), pp. 371– 380
green, 1800–0800 h), four PEV penetration levels (17, 31, 7 Etezadi-Amoli, M., Choma, K., Stefani, J.: ‘Rapid-charge electric-
46 and 62%), five PEV battery sizes and three PEV owner’s vehicle stations’, IEEE Trans. Power Deliv., 2010, 25, (3),
priorities (high, medium and low) are considered. SLM also pp. 1883– 1887
takes into consideration the existing load variations over a 8 Singh, R., Pal, B.C., Jabr, R.A.: ‘Distribution system state
estimation through Gaussian mixture model of the load as pseudo-
24 h cycle while factoring PEV owner preferences for measurement’, IET Proc. Gener. Transm. Distrib., 2010, 4, (1),
charging time zone and priority. Based on this and the load pp. 50–59
flow computed outputs (e.g. voltage profiles and losses), 9 Singh, R., Pal, B.C., Jabr, R.A.: ‘Statistical representation of distribution
SLM schedules when individual PEV chargers should begin system loads using Gaussian mixture model’, IEEE Trans. Power Syst.,
charging such that grid performance is maximised. The 2010, 25, (1), pp. 29–37
10 Singh, R., Pal, B.C., Jabr, R.A.: ‘Choice of estimator for distribution
improvements and benefits of SLM are compared and system state estimation’, IET Proc., Gener. Transm. Distrib., 2009, 3,
demonstrated by performing extensive simulations for a (7), pp. 666– 678
highly detailed 1200 node distribution system topology with 11 Singh, R., Manitsas, E., Pal, B.C.E., Strbac, G.: ‘A recursive Bayesian
several low-voltage residential networks populated with approach for identification of network configuration changes in
distribution system state estimation’, IEEE Trans. Power Syst., 2010,
PEVs. The main conclusions are as follows: 25, (3), pp. 1329–1336
12 Moslehi, K., Kumar, R.: ‘A reliability perspective of the smart grid’,
† The proposed method offers a new and viable integrated IEEE Trans. Smart Grid, 2010, 1, (1), pp. 57–64
means of exercising (semi)automated customer DSM in 13 Civanlar, S., Grainger, J.J.: ‘Volt/var control on distribution systems
smart grids by enhancing DMS capability with SLM, with lateral branches using shunt capacitors and voltage regulators.
Part III: the numerical results’, IEEE Trans. Power Appar. Syst., 1985,
jointly achieving reduced peak demand, improved PAS-104, (11), pp. 3291– 3297
distribution system efficiency and voltage regulation. 14 Denholm, P., Short, W.: ‘An evaluation of utility system
† SLM is shown to successfully maintain voltages within impacts and benefits of optimally dispatched plug-in hybrid electric
tolerances and significantly reduce system losses even under vehicles’. National Renewable Energy Laboratory, NREL/TP-
620– 40293, 2006
high penetration of PEV charging activity. 15 Duvall, M., Knipping, E., Alexander, M.: ‘Environmental assessment of
† A feasible pricing and time zone priority scheme is shown plug-in hybrid electric vehicles’, in ‘Nationwide Greenhouse Gas
to work effectively with SLM PEV charging coordination. Emissions’ (Electric Power Research Institute (EPRI), Palo Alto, CA,
PEV owners can designate preferred charging zones, while 2007, vol. 1)
SLM performs performance improvement functions such as
loss minimisation and peak demand shaving.
† SLM will endeavour to respect PEV owner designated 8 Appendix
charging time zones as long as system constraints (e.g.
voltage regulation and limiting system peak) are not violated. The load and line parameters of the low-voltage 53 node
† SLM is shown to be beneficial in reducing excessive 415 V residential systems (Fig. 5) are listed in Tables 3 and
system loading by controlling peak shaving while also 4, respectively. The parameters of the IEEE 31 bus
reducing the risk of overloads in local circuits including distribution system are adapted from [13].
distribution transformers and cables.
† The impact of PEV was the focus of this study; however, the
proposed SLM approach is applicable to coordinating a wider Table 3 Linear and non-linear (PEV) loads of the typical
range of loads (e.g. PEV charging stations and smart appliances). low-voltage residential system (Fig. 5)
† In our future work, the proposed SLM could be extended with
the integration of new SE algorithms currently being developed Linear and PEV loads Power
to process real-time network data in distribution systems.
Node (Fig. 5) Name kW kVAR
7 References R2 –R53 linear loads 1.5 0.49
selected residential PEV charger 4.0 0
1 Tesla Motors: ‘Tesla motors – high performance electric vehicles’.
2010. Available at: http://www.teslamotors.com/ nodes
2 Nissan: ‘Nissan LEAF electric car’. 2010. Available at: http://www. 2, 4, 7, 9 PEV charging stations 86.4 0
nissanusa.com/leaf-electric-car/

IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877– 888 887
doi: 10.1049/iet-gtd.2010.0574 & The Institution of Engineering and Technology 2011
www.ietdl.org
Table 4 Line parameters of the typical low-voltage residential system (Fig. 5)

Line Line Line Line Line Line


resistance R, V reactance X, V resistance R, V reactance X, V
From node To node From node To node

1 2 0.0415 0.0145 23 24 0.7763 0.0774


2 4 0.0424 0.0189 21 22 0.5977 0.0596
4 6 0.0444 0.0198 16 17 0.1423 0.0496
6 8 0.0369 0.0165 17 18 0.0837 0.0292
8 9 0.0520 0.0232 18 19 0.3124 0.0312
9 12 0.0524 0.0234 16 20 0.0163 0.0062
12 13 0.0005 0.0002 1 35 0.0163 0.0062
12 15 0.2002 0.0200 35 40 0.0415 0.0145
12 14 1.7340 0.1729 40 42 0.0424 0.0189
9 11 0.2607 0.0260 42 44 0.0444 0.0198
9 10 1.3605 0.1357 44 46 0.0369 0.0165
6 7 0.1402 0.0140 46 47 0.0520 0.0232
4 5 0.7763 0.0774 47 50 0.0524 0.0234
2 3 0.5977 0.0596 50 51 0.0005 0.0002
1 16 0.0163 0.0062 50 53 0.2002 0.0200
16 21 0.0415 0.0145 50 52 1.7340 0.1729
21 23 0.0424 0.0189 47 49 0.2607 0.0260
23 25 0.0444 0.0198 47 48 1.3605 0.1357
25 27 0.0369 0.0165 44 45 0.1402 0.0140
27 28 0.0520 0.0232 42 43 0.7763 0.0774
28 31 0.0524 0.0234 40 41 0.5977 0.0596
31 32 0.0005 0.0002 35 36 0.1423 0.0496
31 34 0.2002 0.0200 36 37 0.0837 0.0292
31 33 1.7340 0.1729 37 38 0.3124 0.0312
28 30 0.2607 0.0260 35 39 0.0163 0.0062
28 29 1.3605 0.1357 Distribution transformer reactance 0.0654
25 26 0.1402 0.0140

888 IET Gener. Transm. Distrib., 2011, Vol. 5, Iss. 8, pp. 877 –888
& The Institution of Engineering and Technology 2011 doi: 10.1049/iet-gtd.2010.0574

Вам также может понравиться