Вы находитесь на странице: 1из 7

IEEE Transactions on Power Apparatus and Systems, Vol. PAS-102, No.

5, May 1983 1253


THE IMPACT OF ELECTRIC VEHICLE
DEPLOYMENT ON LOAD MANAGEMENT
STRATEGIES
G. T. Heydt
Senior Member
Purdue Electric Power Center
Purdue University
West Lafayette, IN 47907

Abstract the load management plans will be influenced with


respect to the optimal amount of energy to be shifted
The future penetration of electric vehicles into away from the peak. This influence comes from the
the transportation sector is unknown. However, if off unique characteristics of the EV demand.
peak recharging of electric vehicles is empLoyed, con- This paper considers the impact of EV deployment on
siderable improvement in daily load factor is possi- load management strategies. Many scenarios have been
ble. SimiLarly, load management methods to shift peak studied in EV deployment and references C6-7] further
load to off-peak periods will also improve load fac- document typical studies. The reader should keep in
tor. The impact of electric vehicle deployment on mind that this area is strongLy system dependent and
load management strategies primarily occurs in two numerous parameters have strong influence on conclu-
areas: (i) reduction in effectiveness of load factor sions relative to the net impact of EV's on load
improvement of either electric vehicle deployment or management strategies.
load management strategies due to the action of the
other and (ii) inadvertent addition to peak demand by
electric vehicle load. These and other factors are LOAD FACTOR AND LOAD MANAGEMENT
considered in this paper and several practical con-
siderations of recharging of electric vehicles are If the daily demand profile of an electric power
summarized. system is d(t) watts, 0 < t < 24, with t in hours, the
total daily energy demand in watt-hours is E,
f
ELECTRIC VEHICLE DEPLOYMENT E d(t)dt

Considerable uncertainty surrounds the degree to The peak and minimum demands are denoted with sub-
which electric vehicles (EVs) will penetrate the scripts "max" and "min" respectively. The daily load
transportation sector. Factors which suggest a high factor, P, is
degree of penetration include gasoline price and avai-
lability, potentially lower EV cost due to mass pro-
duction, factors relating to pollution controls, and
= E fT d(t)dt
social acceptability of vehicles of lower performance. T d T d
max max
Each of these factors are uncertain in extent, and
predictions of EV deployment vary considerably. It is where T 24 hours. The sensitivity of fuel costs to
=
clear, however, that no matter what the percent pene- P depend on the incremental energy costs, ac/aw, where
tration of the vehicle market, the electric utilities c and w denote the fuel cost per megawatt hour and
must be prepared to accept this load. Among the fac- instantaneous energy demand.
tors which must be considered are increased loads at If a load management strategy is used, m(t)
the transmission level, increased loads at the sub- megawatts are shifted from the peak to the off-peak
transmission and distribution levels,, and increased period. If the peak period is a < t < b, and the
energy demands. For modest levels of deployment of off-peak period is c < t < d, then
EV's, it is generally believed that increased load
levels will fall well within generation planning lim- m(t) < O a < t < b
its, but there are some concerns about the adequacy of
distribution circuits. Concerning the type of load m(t) > O c < t < d
which the EV presents, factors to be considered in-
clude harmonic signal generation at the EV battery m(t) = 0 otherwise.
charger [1-5), reactive voltampere (Q) demand, battery
charge rates and control, subharmonic signal genera- This is shown pictorially in Figure 1. After load
tion, distribution transformer loading, and voLtage management is employed, the daiLy load factor improves
regulation in the distribution circuit. to P',
Apart from the technical factors mentioned above,
the relationship between EV deployment and load fT (d(t) + m(t))dt
management plans should also be considered. Both EVs
and load management will improve the load factor -- (d max
- m(t peak ))T
but each will do so in a different way. Furthermore,
Consider now the case where
-M a < t < b
m(t) = +M c < t < d
0 otherwise
.82 SM 379-6 A paper recommended and approved by the IEEE then the sensitivity of the'load factor to the managed
Power Systems Engineering Committee of the IEEE Power energy may be calculated,
Engineering Society for presentation at the IEEE PES 1982
Summer Meeting, San Francisco, California, July 18-23, w =fb m(t)dt
1982. Manuscript submitted December 14, 1981; made avail- dP aP . dM
able for printing April 23, 1982. di =
dW '

0018-9510/83/0500-1253$01.00 @ 1983 IEEE


1254
timated and, even if they were both accurate, these
statistical parameters do not determine the probabi;i-
ty density function of the EV demand. Considering
that the EV demand is a non-stationary stochastic pro-
cess with very complex interdependence with other sto-
chastic processes, a lumped model and estimate of the
-n

c
B
variance is probably as well justified as other more
E detailed approaches since the high level of uncertain-
ty does not justify the detail. The lumped model ap-
proach is taken here and the non-stationarity is con-
A sidered at two levels: daily and seasonal variation.
For purposes of simplified organization, seasonal
variation is categorized into two pseudo-stationary
periods: the peak (i.e. Summer and Winter) and the
I - off-peak (i.e., Spring and Autumn).
6 12 18 24 t
Time of Day The active power demand of a battery charges versus
time has a considerable bearing on the impact of EV's
on power systems. Unfortunately, the expression re-
Fig. 1 Typical Load Management Strategy lating active power demand versus time depends on the
battery state of charge and the charger type. The
simplest charger types are uncontrolled rectifiers
dP= 1 Bp with either a series reactor on the ac side and/or a
dW b--a- _J series resistor on the dc side. The function of this
JOd(t)dt reactor or resistor is to limit current. The reactor
BP may be the series leakage reactance of the supply
M (dma xM)2T transformer. Also, the bulk resistance of the rectif-
ier diode may be significant in the total series
Thus resistance of the dc circuit. The average current in
the dc circuit depends on the equivalent resistance of
= 1 d(t)dt the dc circuit. The average current in the dc circuit
T(b-a)(d max -M) depends on the equivalent resistance of the battery
which decreases as the state of charge increases. The
At this point, it is possible to quantify the sen- exact charge characteristics of a lead-acid storage
sitivity of fuel costs to load factor, battery, even under simplified conditions, is a com-
plex function of numerous interdependent phenomena of
dc ac . dW battery state of charge, temperature, and charger
type. In clusters, the active power demand of several
T(b-a)(d max -M)2 a chargers which are activated sometime in the same hour
(for example midnight to 1:00 am) will be rising at a
7-
f d(t)dt decreasing slope until all chargers have been turned
on. Beyond this period, perhaps for 75% of the total
ELECTRIC VEHICLE DEPLOYMENT AND RECHARGE RATES time, a constant or somewhat decreasing charge rate is
maintained to levelize charging and minimize losses.
The energy demand of EVs depend on the following In the later portions of the charge period, a finish-
factors: ing rate at or near the trickle charge level is main-
i. Gross number of vehicles tained. Thus a reasonable charge characteristic ap-
Percent EV penetration, total number of pears in Figure 2 and a total power demand charac-
vehicles in the community, population demography. teristic appears in Figure 3. Reference C8] discusses
ii. Vehicle types a specialized aspect of clusters of EV battery
All electric or hybrid, battery types, chargers, and E93 gives the widely circulated Electric
mechanicaL efficiencies, electrical efficiencies, mo- Power Research Institute (EPRI) recharge scenario. In
tor controller type, battery charger type, vehicle this paper, the recharge rate is considered system
size. dependent, but the characteristic in Figure 3 is used
iii. Battery charge format to obtain typical results. For lead-acid batteries
Battery charging characteristics include ef- and EV's driven about 30Km each day, typical recharge
fects of total energy to be applied to the battery levels are (see Figure 3) as follows for 1000 vehicles
(usually in the range of 0.5 to 1.0 KWH per mile
driven or 4.0 to 0.7 KWH/Km), time of start of charge,
time of end of charge, and charging efficiency.
iv. Use characteristics In
Q
Road types, weather, season, time of week,
holiday occurrence, gasoline price and availability,
economy, employment demographics, other socio-economic 4-4 Battery rated
a)
c
0) t_ cha rge
factors. L.
0.
L DC Charge Current
In each of these areas, considerable uncertainty ex- E
ists and any scenarios formulated are suspect of being 0)
atypical. One approach to formulating EV deployment 0,
L-
n(0
L-
scenarios is to construct several alternative detailed -c
I-,
social-economic-technical structures. A different ap-
L.)

L)

proach is to lump all the multifarious phenomena and 0

estimate a total demand profile. In this approach the T;meI


Time tQ
Q

various sources of uncertainty might be recognized by


estimating a statistical variance for the demand pro-
file. Such an approach is weak in its technical vali- Fig. 2 Simplified Constant Voltage Charge Charac-
dity since the expectation and variance are both es- teristic
1255
Ae
Ae + A a

0)
which increases when energy costs increase (i.e., Ae
AW
40-
Lm ( increases) and decreases when management costs in-
crease. Also, m; = 0 when the load is at the forecast
average load, d. = '. OperationalLy, this formulation
7L uses a forecast for d. and is calculated as the
0) AT H- AT ,.-
I
-i
forecast average. Values of m: are then dispatched to
0~
C -I--t A1 the load management devices (by carrier current or by
_
radio, for example) to operate water heaters, air con-
0 tl t2 t3 t4 t5 ditioners, interruptible loads, etc.
Time (Hours) The disadvantages of this formulation center about
the gross simplicity. Cost functions are assumed to
be quadratic, stationary, and deterministic. Limits
Fig. 3 Active Power Supplied to a Cluster of EV Bat- on m and d are not considered here (although minor
tery Chargers modifications permit inclusion of Limits on m; as ine-
A = 21.4 MW quality constraints). Other costs are ignored such as
B = 42.9 MW maintenance costs, costs due to transmission losses,
C = 4.3 MW. costs relaxed to forced outages, and factors relating
to interchange power.
Table 1 shows a summary of total motor vehicles in For purposes of evaluating the impact of EV's on
the United States C10). It is clear that even a mod- load management, the simplified formuLation given is
est incursion on the total number of motor vehicles in probably adequate considering the high levels of un-
the United States represents a rather large number.
certainty in other salient parameters in the study.
It is estimated that there are already of the order of
10,000 EV's in the United States (1981), and figures SIMULATION RESULTS
of 100,000 to 1,000,000 have been projected over the
next 10 years. One milLion vehicles represents less
than a one percent penetration of the total market. Using the load profile of Figure 4, several load
management studies were run with and without EVs. The
Appendix B presents a brief discussion of the likeli- conditions for all tests were:
hood and influence on load management effectiveness EV Load: Standard Load A
of high penetrations of EV's. EV Energy/Vehicle: 15KWH/(Vehicle day)
Charge Time: 0100 -0600 (24 hour clock)
Table 1 Energy Cost: Figure 5
Total Motor Vehicles in the Load Management Limits: None
United States 10) Load Management Technique: See Appendix
(in millions) Forecast Error: None
Load Profile: Figure 4
Load Factor (Unmanaged base Load,
1977 1978 1979 1982* No EV's): 0.62
Cars 99.9 102.9 104.7 113.0 Case Descriptions A1-A16
.Trucks 28.2 30.5 32.6 39.2
Total 128.1 133.4 137.3 152.2 Table 2 summarizes the several parameters of the
cases studied. In these cases, Ka, the constant which
*Based on a linear trend is proportional to the integral of the square of the
managed power was calculated as proportionaL to the
total cost of the energy in the range in question.
DEPENDENCE OF EV IMPACT ON TYPE OF
LOAD MANAGEMENT STRATEGY
The impact of EV's on load management strategies 2800-
depends on the type of load management used. If a
constant m(t) is empLoyed to shift (b-a)M MWh from the 2400
peak to the off peak, it may be possible to employ the 2000
sensitivity analysis suggested above. If m(t) is
selected in some more compLex fashion, it may be 1l600 \
necessary to resort to simulation to assess the impact
of EV deployment. The Appendix gives a load manage- E 1200 _
ment strategy based on the optimization of a quadratic
energy cost function. This load management may be 800
atypical from practical methods since centralized, op-
timal management is infrequently proposed. On the 400-
other hand, this formulation does capture the desired
functional behavior of managed load and system Load.
As d. increases, m. becomes more negative (i.e., it is 0 2 4 6 8 10 12 14 16 18 20 22 24
desired to reduce the load by Imil to a total of Timp of Day (24 Hour Clock)
mi + d.. Also the amount of management (i.e., size of
mi) is proportional to Fig. 4 Standard Load A
1256
Table 2
Cases Al - A16
Number Load Mgmt Energy of Load Mgmt TotaL Cost Total Cost Load Factor Savings Peak Demand
of EV's Cost/MWMI EV Load Cost Total With Mgmt No Mgmt With Mgmt $(x106) With EV's
Case (MWH)
Number (x1000) (Note 1) (x103) $(x1o6) $(X1o6) $(x106) X (Note 2) MW (Note 3)
Al 5 0.2695 0.075 0.0136 0.97438 1.0087 76.41 0.0239 3000
A2 10 0.2702 0.150 0.0129 0.97642 1.0103 76.50 0.0236 3000
A3 20 0.2714 0.300 0.0101 0.98068 1.0138 76.65 0.0230 3000
A4 50 0.2754 0.750 0.0095 0.99414 1 .0247 77.12 0.0211 3000
A5 100 0.2819 1.500 0.0087 1.01593 1.0435 77.91 0.0189 3000
A6 200 0.2964 3.000 0.0072 1.06017 1.0824 79.43 0.0150 3000
A7 500 0.3446 7.500 0.0056 1.19552 1.2139 83.82 0.0128 3000
A8 999 0.5276 14.985 0.0201 1.45059 1.5020 76.38 0.0313 4007
A9 5 0.1348 0.075 0.0092 0.97326 1.0087 82.93 0.0262 3000
A10 10 0.1351 0.150 0.0091 0.97538 1.0103 82.99 0.0258 3000
All 20 0.1357 0.300 0.0090 0.97967 1.0138 83.12 0.0251 3000
A12 50 0.1377 0.750 0.0085 0.99304 1.0247 83.48 0.0232 3000
A13 100 0.1410 1.500 0.0077 1.01484 1.0435 84.10 0.0210 3000
A14 200 0.1482 3.000 0.0064 1.05943 1.0824 85.26 0.0165 3000
A15 500 0.1723 7.500 0.0049 1.19318 1.2139 88.59 0.0158 3000
A16 999 0.2629 14.985 0.0179 1.43803 1.5020 82.90 0.0462 4007
Note 1: This is Ka factor.
Note 2: Savings is defined as cost without management
minus cost with management.
Note 3: Peak shown is without management.

45.0
0.52 -
40.0 Approximate percent EV
0.48- penet rat ion
35.0 %Il
L
1% 3%b 5% lO0b 30% 50%
0 0.44-
I 30.00 4J
0.40-
3:
I.,. 25.0 0) 0.36- Cases A - A8
9 1
(L)
4fl-0 20.0 CU(o
0.32-
Represent Higher
Load Manage ment
c os t Scenaos rSicea
15.0 L-J

0.28-
10.0 C Cases A 9 - A 1
c) 0.24-
U)c Rep resen t Lowe r|
5.0 CU Load Management/
o~~~~~~~~~~~~~~~~ 0.20- Cos t Scena r ios/
0 co
CU
0 -0 0. 16-
-j
Load (MW) 0. 12
2 IlO 20 lO0 200 1000
Fig. 5 Energy Cost Versus Load Level Number of Ev's (x 1000)

The factor Ka, is the cost of load management. Fig. 6 Load Management Cost for Cases Al - A 16
Thus,
Ka changes from case to case since the range of the
managed power varies with the number of the EV's de- load cost increases due to the higher energy demand.
ployed. Cases Al through A8 represent a higher Therefore, the constant multiplier of
management cost than A9 through A16. Thus these cases m.
are intended to show variation with management costs f 'm(t)dt
as well as EV's deployed. Figure 6 shows K versus 0

the number of EV's deployed. As the EV load in-


a
(mi = managed load) increases to reflect a constant
creases, the managed Load decreases (generally) since ratio with the total energy cost. The realism of such
the total load profile is flatter. However the total a scenario is indeed subject to question, but note
1257
that Ka is not widely variable except for very Large on-peak recharging may cause costs to rise signifi-
deployment of EV's. A more realistic scenario is the cantly and load management might be strongLy advisable
use of a nonlinear function for Ka versus m(t), but to reduce the peak demand. It is concluded that in EV
deployment in which recharging may occur at or near
this would be highly system dependent and also depen- peak periods there may be significant economic motiva-
dent on the types of load management in use. The tion to employ load management. This conclusion ap-
reader is cautioned that this functional variation of plies even for very modest incursions into the tran-
K with managed load is a strong influence in the out- sportation sector (e.g., 1% of total vehicles are
come of a study of the impact of EV's on load manage- EV's).
ment strategies.
The following concLusions are drawn on the basis of LOAD MANAGEMENT TO REDUCE PEAK
these simulations. DISTRIBUTION CIRCUIT LOADING
i. Cases Al - A16 all have the same load profile
(Figure 4 ). The peak demand of 3000MW is represen- In addition to the usual economic motivation for
tative of a community of about 150-300 thousand peopLe load management, Load sites at which EV's are also de-
or about 71-143 thousand vehicLes (using typical auto-
mobiLe data from Table 1). Therefore cases Al - A8 ployed may produce motives related to distribution
represent a penetration of a few percent through circuit peak reduction for purposes of extending dis-
greater than one hundred percent. On the basis of tribution component life. In some areas, distribution
typical automobile data, the first three cases circuits may be of such an age and rating that upgrad-
presented have the most practical significance. The ing would otherwise be required if peak reduction were
higher depLoyment rates are included to indicate not employed. This is illustrated with a simplified
trends, anomalous vehicle/peak MW ratios, and high study of distribution temperature rise considering an
percent penetrations. 8 hour thermal time constant. In this example, ther-
ii. For the conditions stated in Tests Al-A10, the mal overLoad occurs for about 5 1/2 hours per day.
peak demand is insensitive to EV penetration, but the The loading of the distribution transformer exhibits a
minimum demand is very sensitive to EV penetration. linear, first order behavior (Figure 8) for p(t) of
Again, the exact load profiLe will determine how the same profile as in Figure 4. Transformer losses
strong this dependence wilL be.
iii. System load factor will be effected by EV's
onLy above 5% penetrations (typically). Load factor F Z %, Transformer
improvement realizable by management will be effected
only at rather significant EV penetrations -- perhaps ZI- Maximum Temperature-
12% or more. -Temperature No Management
iv. From a purely economic viewpoint, perunitized < cl 110 Rise
savings (savings/totaL) reaLized by load management is
not seriousLy affected by any reasonable EV penetra- ,o K 90 ~ w
tion. Thus, collocation of EV and load management
sites should not reveal any significant economic im- ui 70
pact on a perunitized basis.
vi. Load management techniques measured "per ,J < 3<50 With Management
megawatt-hour shifted" decrease slowly in effective-
ness as EV penetration increases (for 1-5% penetration XJLU 1
30
of the total vehicle sector).
LLJc
In additional studies, several other factors were
considered: on-peak charging, near-peak charging, x: I0 ~Ambient Temperature
Limitations on the load managed, Limitations on the - -

. Hours
uj -1 'a''
energy managed, sensitivity to charge time (i.e., l1 5 9 13
21 17
start of recharge time, end of recharge time, total
recharge time), energy per EV, and alternative battery TIME OF DAY (24 Hour Clock)
recharge characteristics. Figure 7 shows a few con-
sensed results concerning savings realizable by a qua-
dratic formuLated load management method versus EV re- Fig. 8 Temperature Rise of a Distribution Transformer
charge times. The significant conclusion of these with 8 Hour Thermal Time Constant
tests (not shown in Figure 7) is that near-peak or

are assumed to be proportionaL to the square of the


8- load power. If about 10% of the kilowatt hours served
< Savings Realizable By are transferred from the peak period, the peak
A Load Management transformer temperature drops as shown in Figure 8.
.E 6- Algorithm ($ x 10 Per Day) If EV's are permitted to contribute to the peaks, Fig-
4- ure 8 illustrates the potentiaL benefit derivable by a
load management algorithm. References E11-15 docu-
-C ment appropriate analysis' techniques which vary
L 2 Note: Conditions of Tests depending on the type of distribution circuit com-
o A9 - A 16 Apply ponent considered. In the brief example illustrated,
150% transformer loading is assumed to produce maximum
temperature rise in the transformer. The temperature
4-J 0 2 4 6 8 rise plotted is measured in percent of the range
Vn
End of Charge (AM) between ambient and maximum temperature. Thus, the
starting temperature in the day shown is above the am-
bient temperature by 50% of the range between maximum
and ambient. This brief example is intended to em-
Fig. 7 Isopleths Of Savings Realized By A Load phasize that distribution circuit limitations may be
Management Algorithm For Various EV Recharge an important factor in EV deployment in tocalized
Times areas.
1258
Again, it is concluded that EV deployment, if re- ization and Conservation Technology Department
charging at the peak period is permitted, may be sig- Annual Review, 1-3 September, 1981, Washington,
nificant to motivate load management methods to reduce D.C., General Research Corp. Internal Memorandum
peak distribution circuit loading. 2400.
SUMMARY AND CONCLUSIONS [10) U.S. Government Printing Office, "Statistical
Abstract of the United States," 101 ed., 1980
In summary, from the cases described in this paper U.S. Dept. Commerce.
as well as others [16) studied, it is concluded that a
salient factor to be considered in EV deployment is [11) W. Weeks, "Transmission and Distribution of
on-peak or near-peak recharging. This factor should Electrical Energy," Harper and Row Publishers,
be considered particularly in terms of the distribu- New York, 1981.
tion circuit loading. One method to alleviate peak
loading is through the use of load management. Note [123 Westinghouse ELectric Co., "Electrical Transmis-
that the effectiveness of load management, as measured sion and Distribution Reference Book," 4th ed.,
in terms of load factor improvement, may diminish as 1964.
EV's are deployed since this load is likely to improve
load factor and thereby make further improvement less [13) N. E. Chang, "Determination and Evaluation of
warranted. Distribution Transformer Losses of the Electric
The sensitivity of operating costs to load factor System Through Transformer Load Monitoring,"
improvement have also been presented and these sensi- IEEE Trans. Power Apparatus and Systems, V. T-
tivities have been checked for the quadratic- PAS 70, July/August, 1970, pp. 1282-1284.
formulated load management method described.
[14) Y. Bae, P. Adkins, D. Bree Jr., "Quantitative
ACKNOWLEDGEMENT Evaluation of the Cost and Benefits Associated
with Transformer Loading," IEEE Trans. Power Ap-
This work was performed in part for contract paratus and Systems, V. T-PAS 98,
AC02-77CS54250 for The United States Department of En- January/February 1979, pp. 159-164.
ergy.
Mr. L. Conrad, Public Service Indiana, Terre Haute, [15) D. Koval, R. Billinton, "Evaluation of Distribu-
IN suggested study of the distribution circuit load- tion Circuit Reliability," IEEE Trans. Power Ap-
ing. Messrs. J. Mader and F. Klein, Electric Power paratus and Systems, V. T-PAS 98, March/April,
Research Institute, suggested severaL areas of study 1979, pp. 509-518.
and K. Wilson, formerly with the Department of Energy,
served as program manager. Dr. L. L. Ogborn, Purdue [163 G. T. Heydt, "The Impact of Electric VehicLe De-
University, supplied numerous data which come from ployment on Load Management Strategies -- Techn-
practical studies in the EV laboratory at Purdue. ical Issues," Report PCTR 108-81, TR-EE 81-26,
August, 1981, Purdue University, West Lafayette,
REFERENCES IN.
[1] D. Xia, G. Heydt, "Harmonic Power Flow Studies Appendix A
Part I -Formulation and Solution," Submitted A Load Management Strategy Based on
for publication, IEEE Trans. Power Apparatus and a Quadratic Energy Cost Formulation
Systems, 1981.
Consider an energy cost based on fuel cost (cost of
[23 --, "Harmonic Power Flow Studies Part II - Im- generation), Ig, and a cost of load management, Im.
plementation and Practical Application," Loc. Let the total cost include simply these two terms,
Cit.
[33 W. Shepherd, P. Zand, "Energy Flow and Power I =
I9 + Im '
Factor in Nonsinusoidal Circuits," Cambridge
University Press, Cambridge, U.K., 1979. and further let I be a quadratic function of the gen-
eration at time t = i. Let time be discretized (for
[43 R. Owen, M. McGranaghan, J.Vivirito, "Distribu- example, hourly discretization may be used). Also let
tion Systems Harmonics: Controls for Large Power I be a quadratic function of the energy which is
Converters," 1981 Summer Power Meeting, Port- "managed" or "shifted" from one time to another. Then
land, OR, July 1981.
Im is a quadratic function of m;, the power shifted to
[53 W. Shepherd, P. Zakikhani, "Power Factor Correc- time t = i. If time is discretized into hours, m. is
tion in Nonsinusoidal Systems by the Use of both the megawatt power of the shifted load and the
Capacitance," J. App. Phys. D., Vol. 6, No. 16, megawatt-hours energy. If this formulation is placed
26 Oct. 1973, pp. 1850-61. on an incremental basis, the cost I above that for a
[6) R. E. Goodson, "Electric and Hybrid Vehicles Au- uniform generation schedule at 0 Megawatts is approxi-
tomotive Technology Assessment," Office of Tech- mately a quadratic function of the deviation in actual
nology Assessment, January 1980. generation from the mean. Therefore
[7) W. F. Hamilton, Electric Automobiles," McGraw
Hill Book Co., New York, 1979.
9=m
I = Ig+ CI
[83 J. Orr, A. Emanuel, K. Oberg, "Current Harmonics
Generated by a Cluster of Electric Vehicle Bat- = 11 -U)2 +
A e (d.+m EB e(di+m -")
tery Chargers, 1981 Summer Power Meeting,
81-SM-407-6, Portland, OR, July 1981. eC + F2Aam~ + F, B m + Ca
[93 M. M. Collins, "Modeling The Impact of Electric where
Vehicles on Power Generation," EPRI Energy Util- digg = Incremental generation cost above
1259
that for a uniform generating Appendix B
schedule Discussion of Medium and High
c = uniform generating schedule gen- Penetrations of EV's
erating cost
Ae Be" Ce, Aa, Ba' Ca = quadratic cost In this paper, the effects of EV deployment on load
coefficients management effectiveness is examined and an optimal
d. = load power at t load management strategy based on a quadratic cost
formulation is presented. One may reasonably ask for
m. = managed power shifted to time what levels of EV penetrations this analysis is rea-
t = i sonable. When the managed energy is a small percen-
= average load power. tage of the total daily supplied energy, the energy
cost may be approximated by a quadratic function.
Also, the summation is taken over 24 hours to make I Similarly, if the energy supplied to an EV load is
the cost per day. Note that an equality constraint is small compared to the system energy demand in the "EV
recharge period," the quadratic cost formulation is
reasonable. It is concluded that low EV and load
1m; = O managed energy, then, is primarily the case considered
since energy is to be shifted but not altered in to- in this paper. In terms of numbers of EV's, even a
tal. Augment I with a Lagrange multiplier term few hundred megawatt-hours translates into nearly a
million EV's. The subject of this Appendix is the
= I X likelihood and consequences of medium and high pene-
I t cntat, trations of EV's - well above a few percent of the to-
Then to minimize I subject to the stated constraint, tal vehicular population.
The controlling factors relating to EV penetration
a1
in the Long term are largely economic and political.
i 1, 2, 24 Obviously, higher load factors and corresponding sys-
rmm:= O = ...,
tem economies are realizable with any load which has a
significant off-peak demand. Higher EV penetrations
make load management less necessary - this is clearly
seen in the results of the paper. These results are
not sensitive to the type and cost characteristics of
Evaluation of the partials reveaLs, the load management since EV demand is generally a
fraction of the daily minimum load. Opponents and
2Ae (d.+m.- ) + Be + 2A m. + B + 0 proponents of EV's agree that the energy supplied to
EV's is not as significant in influencing penetrations
me = . as economic factors of EV production, battery costs,
public acceptance and commercial promotion. With re-
If the first equation is summed gard to the latter, recent (1982) de-emphasis on Unit-
i = 1, 2, ..., 24, then using 1i = N = 24, ed States Department of Energy programs on EV's sug-
gests that high penetrations of EV's will be well into
2AeNu - 2AeNp + N(Be+Ba) + NX = 0 the future. Note that the Power Utilities Regulatory
Policies Act (PURPA) disallows electric utilities from
Be 8a encouraging any electrical load growth (presumably no
matter what positive benefits might be realizable).
Solving for m;, With a total EV deployment in the United States of
less than 50,000 (less than 0.033%), and in view of
Ae (A-d.) the de-emphasis of EV programs at the governmental
level, it is not likely that the percentage of "elec-
1. Ae +Aa
trics" will increase significantly through 1990.
Beyond this time frame, oil availability and price,
This is the optimal shift in power to minimize the and potential new socio-political atmosphere could
cost as formulated. Note bring about increased interest. Even if EV production
is increased to the point where the population of
11 = -=F, d; electrics double each year after 1990, they will
N
comprise only about 25% of the total vehicular popula-
N = 24 tion by the year 2000 assuming a 2% growth/year in to-
tal vehicles.
The simulation results of the paper for medium and
high penetration levels are still at energy levels
which suggest that the quadratic approximations used
in Appendix A are useable (at least to indicate
trends). For example, if 200,000 EV's exist in a re-
gion with 3000 MW peak demand (estimated population
500,000; estimated vehicular population 380,000; per-
cent penetration of EV's 53%), the EV energy demand is
still less than 15% of the total (assuming 72% load
factor).
It is worthwhile to restate that the EV load, par-
ticularly in high penetrations, is desirable from the
load factor point of view. Evidence of this is the
reduction in necessity to employ load management. On-
ly in the very highest penetrations would peak demand
increase significantly.

Вам также может понравиться