Professional Documents
Culture Documents
Abstract— This paper presents a methodology for modeling range by operating the combustion engine when the battery’s
and controlling the load demand in a residential distribution charge is not sufficient [2].
grid due to plug-in hybrid electric vehicle (PHEV) battery
charging and discharging. To take the stochastic nature of start Nowadays, PHEV are accepted by more and more
charging time, charging during and initial state of charge (SOC) consumers, and its penetration level is rapidly increasing
into consideration, this paper built a stochastic model for PHEV around the world. The estimated number of PHEVs may reach
in a residential distribution grid close to real-world scenarios. 1 million in U.S. by the year 2015 [3] and the penetration level
The authors proposed a smart charging and vehicle-to-grid may reach 60% by the 2050 in U.S. according to the Electric
(V2G) strategy based on particle swarm optimization algorithm. Power Research Institute (EPRI) [4]. The high penetration of
The objective of this control strategy is to improve the power PHEVs in the near future leads to an increase in electrical load
quality and flatten the load demand in the studied system. Then demand. The new load demand will pose a challenge to the
simulations are carried out at different PHEV penetration levels power system especially in small residential distribution grids.
for three different charging scenarios: the uncoordinated In the distribution system, the large load demand caused by
charging, the proposed smart charging without V2G and the PHEVs may increase the peak load, cause the voltage and
proposed smart charging with V2G. The results show that frequency deviation. The PHEVs can also increase the total
uncoordinated charging will seriously increase the peak load harmonic distortion (THD) in the system. The high harmonic
and cause large voltage deviation, while the proposed smart current together with the increased peak load may damage the
charging method can effectively reduce the voltage deviation
transformers in the distribution system. To study the PHEVs’
and flatten the load demand curve. It is found that when V2G is
considered in the proposed smart charging method, the peak
impacts on power system, much research has been done to
load will decrease and the voltage deviation will be smaller too model and control the PHEV load demand. Most studies are
at a low PHEV penetration level, but with the increase of PHEV focused on developing algorithms to control the PHEVs’ load
penetration level, the advantages of V2G will decrease. demand such as sequential quadratic optimization [5], [6] and
dynamic programming [7]. But the PHEV models of [5]-[7]
Index Terms—Plug-in hybrid electric vehicle, stochastic are based on pre-determined charging scenarios which are not
modeling, smart charging, vehicle-to-grid, power quality, very close to real-world situations. References [8]-[10] built
particle swarm optimization. stochastic models for PHEVs but the analysis is limited to
demand modeling. Reference [11] studied the ancillary
I. INTRODUCTION services provided by PHEV through V2G technology such as
The increasing concerns over environment problems and spinning reserves and frequency regulation, but no
energy depletion have brought the development of electric quantitative analysis is included.
vehicles (EVs) into a fast growing stage. According to an To date, seldom studies have been done to develop a
environmental statistics, the transportation sector produced stochastic model for PHEV when power quality issues are
about 29% of all U.S. greenhouse gas emission in 2006 and considered. This study established a stochastic model for
this number is still increasing [1]. The emissions can be PHEV and makes reasonable assumptions close to real-world
reduced by EVs due to fact that the power generated by power scenarios in terms of beginning time of PHEV charging, initial
plants is more efficient than fuel burned in gasoline cars, and state of charge (SOC) and the departure time. This paper also
the emissions of power plants is easier to control. Besides, if applied the V2G technology to improve the power quality in
the EVs are charged by renewable resources, they can achieve the residential distribution grid. The simulations are carried
zero emission. One significant drawback of EVs is the limited out in MATLAB based on a modified PSO algorithm.
drive range. PHEV can overcome the drawback of EV as it
has both electrical and internal combustion engines. PHEV This paper is organized as follows: section II introduces the
has the same advantages of EV when it is operated in all stochastic model for PHEV, section III discusses the
electrical mode and it can satisfy the drivers with large drive mathematical model for PHEV charging and proposes a PSO
algorithm based control strategy, section IV introduces the
,((( 120
case studies and simulation results, and section V draws The distribution of the arriving PHEVs can be expressed as
conclusions. normal distributions as shown in Fig. 2.
0.45
Probability density
0.3
The charging demand of a PHEV is associated with many
0.25
factors, such as initial battery SOC, charging power, the 0.2
charging start time and the plug-in duration. This study is 0.15
based on a residential distribution grid environment, so the 0.1
PHEVs will only be plugged into the grid once a day when 0.05
immediately after they come home at night and the PHEV Figure 2. Probability distribution of arriving time of PHEVs.
will be fully charged when they leave home the next morning.
The PHEVs are assumed to be operated in all electrical mode The PDF of the arriving time of PHEVs is as follows:
until the battery SOC reaches to 20% to extend the battery 1 మ
life. ܨ் (= )ݐ ݁ ି(௧ିఓ)Τଶఙ , 0 < < ݐ24 (3)
ߪξ2ߨ
The initial SOC of a PHEV can be derived from its daily where t is the time in a day, μ=18, ߪ = 1
travel distance. According to the US National Household Also, the distribution of departure time of PHEVs in the
Travel Survey (NHTS) [12], the distribution of travel distance next morning can be expressed as:
can be described by a lognormal distribution as shown in Fig. 1 మ
1. This distribution has zero probability for negative values of ܨ் (= )ݐ ݁ ି(௧ିఓ)Τଶఙ , 0 < < ݐ24 (4)
travel distance and a low probability for long travel distance. ߪξ2ߨ
where μ=7, ߪ = 1
According to Fig. 1, most travelers have a daily travel range
As the arriving and departure time of a PHEV can be
between 5 to 20 miles.
obtained from (3) and (4), the plug-in duration could also be
0.06 calculated.
0.05
B. Vehicle to Grid
The basic concept of V2G is that the PHEV can provide
Probability density
0.04 power to the grid while it is plugged in. Usually, the PHEV
can charge during off peak time and discharge when power is
0.03
needed by the grid. In this study, the V2G technology is used
0.02 to shave the peak load demand.
III. PROBLEM FORMULATION
0.01
121
The basic principle of the proposed smart changing is to maximizing the load factor should have a good effect on
manage the load demand of PHEVs to improved load factor reducing voltage deviation. Also, mazimizing load factor will
of the system by charging at off peak times. In order to shift obviously flatten the load demand. Thus, the objective
more loads from peak times to off peak times, the charging function is to maximize the load factor:
power of PHEVs should be set to the rated power. Then the
charging process of a PHEV can be designed as a charging ܦ௩
sequence as follows: max{ (15)
݉ܽݔ൫σௗ௦
ୀଵ ܦ,௧ ൯
் where
ܸ is the system input voltage;
ܴெ · ߩ ܧ = )ݐ(ݎ, = )ݐ(ݎ0,1 (6)
ܸ is the voltage at load point ݊;
௧ୀଵ
When V2G is considered, then: ܦ,௧ is the load demand of load point n at time ;ݐ
ܦ௩ is the average of the total system load demand;
் ் ܴ is the resistance of line ݈;
ܴெ ή ߩ න ݎ ( ݐ݀ )ݐ+ ܴெ න ݎௗ (ܧ = ݐ݀ )ݐ (7) ܫ,௧ is the line current of line ݈ at time ݐ.
where C. Particle Swarm Optimization
ܴெ is the maximum possible charging power; A particle swarm optimizer is developed to solve the
ܴெ is the maximum possible discharging power; formulated problem. The basic principle of PSO is to find the
ݎ ( )ݐis the charging rate at time t, 0 )ݐ(ݎ 1; best position of a particle in order to optimize the objective
ݎௗ ( )ݐis the discharging rate at time t, െ1 )ݐ(ݎ 0; function in the search space.
For simplicity, assume: The PSO algorithm [13] has many particles in the system
and every particle has its own location and velocity. These
ܴெ ή ߩ = ܴெ = ܴெ (8) particles can learn from their own experiences and the
ݎ ( )ݐ+ ݎௗ ()ݐ(ݎ = )ݐ, െ1 )ݐ(ݎ 1 (9) experiences from other particles. The position and velocity of
each particle are adjusted according to equations (15)-(17).
So equation (7) can be rewritten as: The velocity of a particle is affected by the its own best
் position ݐݏ݁ܤand best position of all the particles ݃ݐݏ݁ܤ.
ܴெ න ܧ = ݐ݀ )ݐ(ݎ, െ1 )ݐ(ݎ 1 (10)
ାଵ
ݒௗ = ݒݓௗ + ܥଵ ή ݀݊ܽݎଵ · ൫ ݐݏ݁ܤെ ݔௗ ൯ + ܥଶ ή ݀݊ܽݎଶ
Similarly, the charging and discharging power of PHEV
· ൫݃ ݐݏ݁ܤെ ݔௗ ൯ (15)
should be as large as possible to shift more charging loads to
the off peak times and provide more power to the grid during
ାଵ ାଵ
peak hours. The charging process can be designed as a ݔௗ = ݔௗ + ݒௗ (16)
charging sequence as follows:
ݓ௫ െ ݓ
் ݓ = ݓ௫ െ ݇ ή (17)
݇௫
ܴெ ܧ = )ݐ(ݎ, = )ݐ(ݎെ1,1 (11) where
௧ୀଵ
subject to (12)-(14): ݒௗ is the velocity of particle ݅ at dimension ݀;
௧భ
ݔௗ is the position of particle ݅ at dimension ݀;
ܱܵݐ(ܥଵ ) = (ܱܵ(ܥ0) ή ܧெ + ܴெ ) )ݐ(ݎ/ܧெ (12)
௧ୀଵ
ݓ is the inertia weight;
0.2 ܱܵݐ(ܥଵ ) 1 (13) ݇ is the iteration number;
0 ݐଵ ܶ (14)
where ܥଵ and ܥଶ are the learning factors.
ܶ is the expected plug-in time; In the PSO algorithm, the fitness value of each particle is
ܧெ is the battery capacity; calculated according to the objective function at each
ܱܵݐ(ܥଵ ) is the battery state of charge at time ݐଵ . iteration. When a best fitness value is found for a specific
According to equation (11), the charging process of a particle, the position of this particle will be the ݐݏ݁ܤvalue
PHEV can be simplified as a charging sequence based on for this particle. Similarly, when a best fitness value for all
each hour. The charging sequence consists of “1” and “-1”, the particles is found, the position of the particle will be the
where “1” means charging at this specific hour and “-1” ݃ ݐݏ݁ܤvalue. After all iterations are over, the best position
indicates discharging. ݃ ݐݏ݁ܤcan be found in the searching space.
B. Tested objective functions In this specific problem, we need to find the optimal
In this study, the basic goal to control the sequences of charging and discharging sequence of each PHEV in the
PHEVs is to achieve better power quality and flattened load system. So the location of a particle in the search space should
include the information on the sequence of each PHEV. The
demand. As voltage deviation is due to the system over load,
number of PHEVs is set to be the dimensions of the search
122
space of each particle and the number of possible charging and departure time. Then the aggregator can control the PHEVs
discharging sequences of PHEV is the searching range of the based on its own algorithm. In this study, the aggregator is
dimension. For example, if a PHEV will be plugged in for 9 designd to improve the power quality of the system by
hours and needs 5 hours to be charged to full SOC, it will has minimizing the load voltage deviation and flattening the
7 hours for charging and 2 hours for discharging when V2G is system load demand. As the a basic performance index of an
considered. One possible charging sequence could be “1-1- aggregator, the revenue will also be considered in this study.
1111111”, and the number of the total possible charging The revenue is caculated based on the varying electricity
sequences for the PHEV is ܥଽଶ = 36. Then this 36 different price. Fig. 5 is an locational marginal pricing (LMP) from
sequences are encode as 1 to 36 as the searching range for this PJM [17]. Although the price curves are not the same
dimension.
everyday, the daily pattern is similar. Fig. 5 is a typical price
IV. CASE STUDIES curve and it is used in this study to generate charging cost in
A. The Test System the system for differnent control strategies.
60
1400
According to Fig. 6 and Fig. 7, at low PHEV penetration
levels, the smart charging has a better performance on
Power(W)
1200
400
penetration level increases, the advantage of V2G is reduced.
200 This phenomenon can be explained by the following analysis:
0
0 5 10 15 20
When PHEV penetration level is low, after the aggregator
Time(h)
allocates PHEV charging load to the off peak time, there is
still a large gap between peak load and off peak load, so the
Figure 4. Household load during winter [15].
V2G strategy can still shave the peak load as shown in Fig. 7
An aggregator is in place in the residental distribution grid (a)-(c). But when the PHEV penetration level is very high,
to manage the charging process of the PHEVs in the system. after aggregator flattens the load demand, there is no peak
The aggregator makes contracts with vehicle owners, so once load demand existing. The V2G strategy can not make the
a PHEV is plugged into the system, the vehicle owner should load demand more flattened. So the V2G strategy exhibits a
provide the information about arriving time, initial SOC and better performance at low PHEV penetration levels.
123
Uncoordinated charging Uncoordinated charging
Smart charging without V2G Smart charging without V2G
Smart charging with V2G Smart charging with V2G
Time (hours) Time (hours)
a. Load voltage curves at 10% PHEV penetration b. Load voltage curves at 20% PHEV penetration
Uncoordinated charging Uncoordinated charging
Smart charging without V2G Smart charging without V2G
Smart charging with V2G Smart charging with V2G
Time (hours) Time (hours)
c. Load voltage curves at 40% PHEV penetration d. Load voltage curves at 60% PHEV penetration
Figure 6. Load votage curves for node point 34 of the residentail distribution grid at different PHEV penetration levels.
Base Load Base Load
Uncoordinate charging
Uncoordinate charging
Smart charging without V2G Smart charging without V2G
Smart charging with V2G Smart charging with V2G
Load Demand (KW)
Time (hours) Time (hours)
a. Load demand curves at 10% PHEV penetration b. Load demand curves at 20% PHEV penetration
Time (hours) Time (hours)
c. Load demand curves at 40% PHEV penetration d. Load demand curves at 60% PHEV penetration
Figure 7. Total load demand of residential distribution grid at different PHEV penetration levels.
124
actually costs more money than the smart charging
Uncoordinate charging
strategy without V2G.
Smart charging without V2G
Cost of Charging ($)
V. CONCLUSION
Smart charging with V2G
This study presents a methodology to integrate PHEVs
in a residential distribution grid. The paper builds a
stochastic model of PHEVs in residential distribution grid.
According to this model the authors proposed a smart
charging and V2G method based on a modified PSO
algorithm to improve the power quality in the studied
system. The authors study the performance of the proposed
smart charging algorithm with and without V2G
technology at different PHEV penetration levels. The
PHEV Penetration Level simulation results show that when PHEVs are charged
Figure 8. Costs of charging by different control stategies. without control, there will be a very large deviation on load
voltage, but when smart charging is applied, the voltage
Fig. 8 shows the charging cost of PHEVs based on deviation is reduced, load demand profile is flattened and
different control strategies. Compared to uncoordinated charging cost is reduced. The simulation also proved the
charging, the smart charging strategies reduce the cost V2G strategy exhibits a better performance at low PHEV
effectively by avoiding charging at peak hours. When penetration levels. When the PHEV penetration level
PHEV penetration level is low, the V2G stategy saves increases, the advantages of V2G technology will decrease.
more money than the smart charging stategy without V2G
by selling electricity back to the grid at peak hours. When ACKNOWLEDGMENT
PHEV penetration level is higher than 40%, the V2G This work is in part supported by California Energy
stategy costs more money than smart charging stategy Commission through the Energy Innovations Small Grant
without V2G as the peak load does not exist after smart (EISG) Program.
charging as shown in Fig. 7(c), (d). In this situation, as a
portion of energy is wasted in the converter during the
charing and discharging process of V2G, the V2G strategy
125