Professional Documents
Culture Documents
Abstract-In order to accurately describe the influence of the permeability of wind power and electric vehicles in the
random factors in distribution systems on the probabilistic load distribution system is bound to bring the impacts and
flow (PLF), a PLF method of distribution system containing challenges in aspects of voltage modulation, frequency
wind power and EVCL based on a stochastic response surface modulation as weil as security and stability of the distribution
method (SRSM) is put forward. The proposed method can system. Therefore, considering the output randomness ofwind
sufficiently consider the randomness of wind speed as weil as power and the uncertainty of electric vehicles performs PLF
uncertainty of basic load and EVCL. Furthermore, the analysis, which has great significance in aspects of economy,
orthogonal transformation is introduced to deal with the
security and stable operation in the system [2].
correlated wind speeds among different wind farms. Comparing
PLF is an effective method for the quantitative analysis
with the traditional Monte Carlo simulation method with simple
of the above-mentioned problems, which can model the
random sampling (MSRS), the proposed method can solve
problems that they are facing, Iike more number of calculation uncertainties as input random variables with probabilistic
times, lang simulation time. Simulation resuIts on IEEE14 and density functions (POF) and cumulative density functions
IEEE30 bus system demonstrate the validity, accuracy, rapidity (CDF) [2]. Recently, the PLF methods have been extensively
and practicability of proposed method. In contrast to the MSRS, used to analyze the distribution network operation under
the proposed method is of higher computational efficiency and uncertainties and to evaluate the impact of renewable energy
beUer simulation accuracy. resources. In Ref. [3], the impact of wind generation on
distribution system has been studied using PLF. In Ref. [4],
Index Terms--EVCL; orthogonal transformation; PLF; SRSM; the electric vehicles with stochastic charging behavior have
wind speed correlation. been also modeled and evaluated as a PLF problem. Ref. [5]
has introduced PLF to perform analysis of the power system
I. INTRODUCTION containing wind power generation, photovoltaic power
generation and electric vehicle. In Ref. [6], the PLF of the
Since the fossil energy is gradually exhausted and the power system containing wind turbine and photovoltaic cell
problem of global warming caused by greenhouse gases is system has been calculated with the adoption of LHS method
serious by degrees, in recent years, wind power, as an and Nataf transformation. In Ref. [7], the influence of wind
important part of the renewable energy, has been rapidly speed correlation on transient stability of power system has
developed all over the world. However, the uncertainty of been studied by using SRSM. In Ref. [8], the probability
wind power, which is caused by the randomness of wind model containing wind farm has been adopted to carry out
speed, has also brought aseries of problems when the large- multi-objective reactive power optimization by using SRSM.
scale wind power is integrated into the existing power grid [1]. Based on the research above, the accuracy and rapidity of
In addition, the emerging more active and stochastic new load SRSM have been demonstrated.
such as electric vehicles have had the incomparable The wind speeds among different wind farms are
advantages of conventional cars in the aspects of the appreciably correlated in close proximity, which has a
alleviating energy crisis and energy conservation and significant impact on the planning and operation of power
emissions reduction, there are more and more uncertainties system, hence it is of great significance to consider correlation
being introduced in distribution systems. Hence the integration factors in order to assess load flow characteristics of power
of wind power and electric vehicles has made the operation system accurately and comprehensively. Recently, the
and control of the power distribution system to face more orthogonal transformation has been widely applied in the PLF
uncertainty [2]. Properly dealing with the uncertainties has calculation with correlated input random variables. In Ref. [9],
become a practical challenge. The gradually increasing
1861
the extended point estimation method based on orthogonal N=(m+n)! (4)
trans form has been researched, which has been applied to m!n!
calculate the PLF of wind farm with wind speed corre1ation. The key to determine the undetermined coefficients
In Ref. [10], the corre1ated wind speed has been modeled by
ao' all ,a!l!' ,... is the se1ection of collocation point. In order to
Nataf transform, and then the transient stability based on
SRSM has been analyzed and studied. ensure the computational efficiency and simulation accuracy,
Based on the above analyses, a PLF method of distribution the optimal collocation points are selected according to linear
system containing wind power and EYCL based on SRSM is independent principle.
put forward. The optimal collocation points are selected C. Selection ofpolynomial expansion term based on the
according to linear independent principle, and the stepwise stepwise regression analysis
regression analysis is introduced to screen out minor items in According to the stepwise regression analysis, the flow of
response surface expansion terms. This method sufficiently selection of polynomial expansion term is shown in Figure 1,
considers the randomness ofwind speed as weil as uncertainty the significant level of impact of the expansion term of
of basic load and EVCL, and the orthogonal transformation is polynomial on the output random response is expressed by
introduced to deal with the corre1ated wind speeds among
inspection value F; ofthis expansion term.
different wind farms, thus PLF in which the correlation of
Start
wind speed is taken into account can be calculated, the
influencing variation law of wind speed correlation on PLF as Input the llumber of random variables and order of polYllomial
chaos expansion, and calculate the llumberofulldetermined
well as the configuration of reactive power compensation coefficients of the Hertnite 01 nonrial chaos ex ansian
capacity at grid-connecting point of wind farms under the
Accorcli ng to 1i TIeaf independent principle, seleet the optim aI co Ilocati on poi nt
consideration of wind speed corre1ation are researched and
analyzed. Simulation results on IEEE14 and IEEE30 bus Accordillg to the collocation point, calculate
system have demonstrated the validity, accuracy, rapidity and Hennite 01 lOmial alld Oll ut 5tO chastic res onse
F; = p;(n-m-l)/S; (5)
(2) /1 /1
1862
D. Characteristics ofprobability and statistics ofthe output The relationship between the power output and wind speed
random response meets the power characteristic curve ofwind turbine [5].
To determine the undetermined coefficients of polynomial
chaos expansion by power tlow calculation of sampie points,
and the mean value and standard deviation for the output
random responses are obtained by Monte Carlo method.
Pwind = j O'
a+bv,
Pr.
(9)
0,
III. ORTHOGONAL TRANSFORMATION
where vci ' vco and vr are cut-in wind speed, cut-out wind
When the input random variables are not independent, the
intluence of the correlation between random variables on the speed and rated wind speed. p, is the rated active power of
PLF calculation must be taken into account. Orthogonal wind turbine. P'rind is random output ofwind turbine generator.
transformation is adopted to transform the correlative input a = Pr V ci /(v ci - vr) , b = p,. /( VI' - V Ci )
random variables into uncorrelated random variables. To the
B. Probability distribution model of E VCL
random variable X = [xpxz, ... ,xnf, which is composed of n
EVCL characteristics are mainly affected by the charging
random variables, the mean value of each random variable
number, the charging rate, the charging start time and the
is P = [ppPz, ... ,pS, the standard deviation of each random charging duration. The daily mileage of vehicle is
variable is 0'= [O'P0'2'''''O'nf, the correlation coefficient lognormally distributed, whose PDF is
between Xi and Xi is Pu ' the expression of covariance matrix is J(d) = 1 exp[ (lnd - ~ld )2] (10)
2 dO"d& 20"d
0'1 P120'10' 2 P1n O'lO'n where d is daily mileage, Pd =3.2, O'd = 0.88.
2
Cx-- P120' 20'1 0'2 P 2n O' 20' n Eis defined as starting value of electric vehicle battery's
SOC, E = (1- diR) , R is mileage range of electric vehicle after
2 a full charge, the value of R is 100kM. Combining with
P1I1 0'nO'j PZ n O'n 0' Z O'n
where C x is the symmetric matrix, there is an existing formula (10), whose PDF is
J( E ) = 1 [ {ln[R(1 - E] - ,LId } 2 ]
orthogonal transform matrix B , the random variable X is exp 2 (ll)
transformed into uncorrelated random variable Z . R(1- E)O"d& 20"d
Z = BX (6) Start charging time are normally distributed, its PDF is
where the covariance matrix Cz of random variable Z is the 1 (1_j1,)2
1
~ exp [-.- - 2- ] , (1', -12) < 1.:0: 24
identity matrix, namely, C z = I . <7 ,,2Jr 2<7, (12)
j~)= ' ,
Cx is symmetric positive definite matrices, which is I (t + 24 - pY
~ exp[ , ] , 0 < 1:0: (1', -12),
adopted to carry out Cholesky decomposition, namely, <7,,,2ff 2<7;
C x = LL1 where PI =17.6, 0'1 =3.4.
According to formula (6) can be derived: ~. is charging power, tEV is charging duration, PDF is
C z = f co v(Z,Zl) = fCOy(BX,X1Bl) = Bfco v(X,X 1)B T J( ~v ) -- 1
1
[3l.61xO.15
x
= BC xBT = B(LLT)B T = I => B = L- 1 (7) 1.61xO.15 2 tFVO"d&
In the PLF calculation and analysis base on SRSM, it is exp[ (lntFv-lnI.61-1n~.15+1np-,LId)2 ]dp (13)
necessary to transform the correlative random variable X into 20"d
uncorrelated random variable Z . There are N electric vehicles in the system, the daily
mileage of vehicle d and the charging start time t are mutually
IV. PROBABILITY DISTRIBUTION MODEL FOR WIND POWER independent, then at the t(t ~ T) moment, the expected value of
ANDEVCL
active demand for electric vehicle is
A. Probability distribution model ofwind power P(t, T) = N ['J(E)p(t)dE (14)
Since the uncertainty of wind power generation is directly Jo
related to the uncertainty of wind speed, it's important to The expression of pet) is
1
select the appropriate model ofwind speed, analysis of a large
Pe.,1 E [T,T + I El' ]
number of measured data shows that the wind speed in most , T + t EV :0: 24
areas obeys double parameter Weibull distribution [5], PDF of
wind speed for wind farm is
p(t) =
1{
0, I E [(0, T) u (T + I El' ,24 )]
Pol E [(O,T + I KV -24)u[(T,24)]]
, T + t vv 2: 24
K . V -' V K
(8) O,IE[T+I KV -24,T]
j(v)=-(-) exp[-(-)]
C C C tEv = (1- E)I Pe (15)
where K is shape parameter, K = (0" / ,LIr 1086 C is scale The expected values of active demand for N electric
parameter, C =,LI 1[['(1 + 11 K)], P is average wind speed, (J' is vehicles are obtained at t moment from the formula (15).
standard deviation. [' is Gamma function. Combining with central limit theorem, EVCL approximately
1863
obeys the normal distribution at t time. On the basis of the TABLEI. STATISTICS OF VOLTAGE FLUCTUATION OF 29 BUS
UNDER DIFFERENT CONDITIONS
known mean value JI,iV and standard deviation (J",y , thus
Conditions P(V<O I maximum Minimum mean standard
EVCL'sPDFis .951 voltae:e/v.u voltae:e/p.u. value deviation
no
0,1275 1.0052 0,9273 0,9731 Om02
= __l_exp[_ (PFV - ~IFV )2 1
strong
f(Ptf) (16) POSlttV
fluctuation
10010
.j2;O"tf' 20"FV e
fluctuation
0,1312 1.0197 0,9156 0,9735 Om09
correla
20%
bon 0,1390 0,0278 0,9072 0,9732 Om12
V. PROCEDURE OF PLF CALCULA TlON BASED ON SRSM fluctuation
no
AND ORTHOGONAL TRANSFORMATION moder fluctuation
0,0728 1.0056 0,9283 0,9738 0,0091
ate 10%
The wind speed and the power ofthe basic load and EVCL correla fluctuation
0,0861 1.0144 0,9145 0,9736 0,0108
1864
random tluctuation of load is considered and the installed .3 ..S
capacity of wind farm is different. Therefore, the reactive
power compensation capacity that should be configured 2 ..S
0.1
According to (c), (d) in the Figure 4, the access of wind
s 10 1.5 20 25 30 farm has had a great intluence on the branch power tlow, and
standard deviation of load fluc t uation
it has also changed the tide of the original tlows. The
Figure 2. Average and maximal ARMS ofbus voltage amplitude
probability for power tlow overloads ofbranch 27-30 has been
The simulation results have shown that SRSM is of higher increased from 0 to 6.73%.It may be concluded that different
computational accuracy when the load tluctuation is lines has different intluences affected by the wind farms'
exceedingly smalI, the calculation accuracy of SRSM integration. The more closer to the integration node, and the
gradually decreases with the increase of load tluctuation. more heavier the original power courses, the more severity of
Case2: a wind farm is installed at node 30, Pr =IMW, line tlow's tluctuation will be got.
K =3.97, C =10.7, V ci =3m/s, V r =12 m/s, V co =25 m/s, the
Case3: EVCLs are installed at node 30 on the basis of
Case 1. There is the electric vehicle battery with the capacity of
standard deviation of the loads equals to 10% of their 18kW . h, charging power of the single electric vehicle is
expectation. The relationship between the average and 3.6kW, when a battery charges, conversion efficiency is 0.75.
maximal ARMS of bus voltage amplitude and the installed The relationship between average and maximal ARMS of bus
capacity ofwind farm is shown in Figure 3. voItage amplitude and the number of EVCLs is shown in
Figure 5.
1865
active
0.237 2.263 0.575 3.059
power
O. R-
reactive
0.529 1.597 0.736 2.986
f O. 6- power
~O ..1- "alup
From Tab.7, the maximum value of s;mcan is 2.401 %, the
0. __ ________ _ maximum value of Ei;",,", is 4.787%. It is seen that the relative
O'c-----c2ccOOc---cc40CC'O------c~---c8~OOcc--lcc!OOO error index for each output random variable is all less than 5%.
thp llnmher
Hence one can see that the error of SRSM is smaller, which
Figure 5, Average and maximal ARMS ofbus voltage amplitude can be used to accurately evaluate the probabilistic
The simulation results have shown that the average and characteristics ofsystem containing wind power and EVCL.
maximal ARMS gradually increase with the increase of the In the case study, the calculation time of MSRS is 68.95s,
number of EVCLs. and yet that of SRSM is 1.67s, only about 2.4% of MSRS,
Case4: EVCLs are instalIed at node 30 on the basis of which has shown that SRSM has the advantage of fast
Case2, there is the wind farm operated in constant power calculation speed in solving PLF.
factor with the capacity of 5MW, the number of EVCLs are VII. CONCLUSIONS
600. The voltage ofnode 30 and the power flow ofbranch 27-
1) The load flow of distribution system has been affected by
30 are used as the test point, which are shown in TABLE ITT.
the output randomness of wind power and the uncertainty
TABLEIII. RESULTS OF PLF CALCULATION ofbasic load and EVCL, significantly.
standard 2) After considering the correlation ofwind speed, the wind
Case mean value
standard mean value deviation speed correlation has a great influence on the results of
deviation of reactive of reactive PLF.
cIassific ofvoltage for
ofvoltage power for power for
ation 30 bus
for 30 bus 27-30 branch 27-30
3) The reactive power compensation capacity of grid-
branch connected point for each wind farm has also changed
Casel 0,9947 0.0050 -0.0082 0.0344 when the instalIed capacity of wind farm is different.
Case2 1.0061 0.0127 0.0186 0.0447 4) The simulation accuracy and computational efficiency of
Case3 0.9969 0.0129 0.0193 0.0432 SRSM have been better than MSRS, which has been able
Case4 1,0075 0.0128 0,0188 0.0429 to accurately evaluate power system operation and
The result obtamed by MSRS wlth the sampIe slze of analysis containing wind power and EVCL and has had a
10000 is set as reference, SRSM and MSRS with the sampie good prospect in engineering application.
size of 500 are adopted to carry out PLF calculation under the
Case4, the average and maximal ARMS of voltage amplitude, REFERENCES
active power and reactive power are shown in TABLE IV. [I] Thomas Ackennann, "Wind power in power systems," Wiley, 2012,
TABLEIV. AVERAGE AND MAXIMAL ARMS OF VOLTAGE AND
[2] Villanueva D, Pazos J L, "Probabilistic load flow including wind power
UNE FLOW generation," IEEE Transactions on Power Systems, vol. 26 (3), pp.
1659-1667,2011,
index value/lO-2 index value/lO-2 [3] p, Caramia, G, Carpinelli, M. Pagano, "Probabilistic three-phase load
Index Index
SRSM MSRS SRSM MSRS flow for unbalanced electrical distribution systems with wind farms,"
AUtf 0,89 1.50 AUJ 0,72 1.21 lET, vol. 1(2),2007.
Al'M 0.67 1.31 Al'A 0.46 1.03 [4] G. Li, X. p, Zhang, "Modeling of plug-in hybrid electric vehicle
AU\.f 1.75 2.52 Au! 1.51 1.98 charging demand in probabilistic power flow calculations," IEEE
Trans-actions on, vol. 3(1), pp. 492 -499, 2012.
From Tab.6, ARMS values obtamed by SRSM are mostly [5] C. X. Wu, F. S Wen, "Probabilistic load tlow of power system with
less than 0.02. Therefore, the results obtained by SRSM are WFs, PVs and PEVs," Electric Power Automation Equipment. vol.
dose to that of MSRS with sampie size of 10000, comparing 33(10), pp. 8-15, 2013.
with the resuIts obtained by MSRS with sampIe size of 500, [6] Y. Chen, J Y. Wen, S. J Cheng, "Probabilistic load t10w method based
calculation accuracy of SRSM has been effectively improved. on Nataf transformation and Latin hypercube sampling," IEEE Trans
on Sustainable Energy, vol. 4(2), pp, 294-301,2013.
The average value &:mean and maximum value s;mox for [7] L. Zhang, J D. Huang, "Impact of wind speed correlation on transient
output random variables are employed to evaluate accuracy of stability of system," Power System Proteetion and Control, vol. 42(6),
pp. 77-83,2014.
the proposed method, which are shown in T ABLE V. [8] Z.G, Liu, H. Liu, .I. Liu, "Multi-objective reactive power optimization
TABLE V. TAB.7 RELATIVE ERROR INDICES OF OUTPUT considering wind farm probabilistic model," Power System Proteetion
VARIABLES andControl, vol. 41(1), pp. 197-203,2013.
[9] M. Fan, V, VITTAL, G. T. HEYDT, "Probabilistic power flow studies
y
C.;mean /0/0 C. Sffi!1X 1% for transmission systems with PV generation using cumulant," IEEE
Output
variables standard standard Trans on Power Systems, vol. 27(4), pp. 2251-2261, 2012.
mean value mean value
deviation deviation [10] X. Pan, W. X. Liu, Y. Q. Xu, "Transient stability analysis of power
voltage system integrated with wind farm groups based on SRSM and Nataf
0.014 1.816 0.037 3.078
amplitude method, "Proceedings olthe CSEE, vol. 33(16), pp. 56-62,2013.
voltage [11] X. F. Wang, "Power system optimal planning," Beijing Hydraulic and
0,156 2.401 0,178 4,787
phase Electric Power Press, 138-148,321-32,1990,
1866