Professional Documents
Culture Documents
ed
min f (u,x) (1) s.t. g(u,x) 0 h(u, x) < 0 where u is the decision vector, consisting of all control variables such as active output of generators and the terminal
1-4244-0111-9/06/$20.00(c2006 IEEE.
voltage magnitudes of generators. x is the state vector, consisting of the voltage magnitudes and phase angles of all the load buses and reactive power output of every generators in the system. f is the objective function of grid active losses which is the sum of losses on each branches. g denotes the power flow equations. h denotes system operating limits of state vector, decision vector and other inequality constraints as well. The nodes in power system usually can be categorized into PV node, PQ node and slack node. But wind generator dose not act as these three kinds of nodes, which makes the power flow equation g(u,x)=0 become much more complicated to solve. Some scholars have dealt with wind generator as an approximate PQ node. However, the reactive power output of wind generators is a variable dependent on the active power, terminal voltage and rotor slip. A typical equivalent circuit of induction generator[ 0] can demonstrate the relations clearly. Fig. 1 shows a simplified equivalent model of induction generator. In Fig.1 and the following paragraphs, x1 is stator reactance, x2 is rotor reactance, r2 is rotor resistance, xm is excitation reactance, s is rotor slip, U is terminal voltage, I is terminal current, P is active power output and Q is reactive
power
r2 +Xk (Xk+Xm)s p
r2XmS
(5)
output.
iX I X, 5g A.
According to the simplified equivalent model, active power output is described as follows:
p
(r2
U2r2 IS
s)2 +x2
(2)
VU4r2 - 4P2x2r
2Px
2
arctanK
r2
+ Xk(Xk+ Xm
r2X,S
(4)
Based on (2) to (5), it is easy to follow that reactive power output Q is affected by active output P and terminal voltage U. The active output is constant at given wind speed, but the terminal voltage is determined by the power flow of whole system including wind generators, which is right the difficulty in solving power flow equations. If Q is also taken as constant value, the wind generator node becomes a PQ node, calculation of power flow equations is just like that in conventional system, but the solution is only an approximation to the real one. If take account of (5) in power flow calculation, the process becomes very computational, wind generator equation and the rest power flow equations are calculated by turns, for several iterations a compromised solution can be achieved. This is called iteration model in this paper. And the former one is called PQ model. In calculation of ORPF, PQ model may bring more warps, but the iteration model definitely brings terrible large quantity of computation. Besides the two models, a new idea is proposed to cope with wind generators in system optimal problems. From the system viewpoint, wind generation only occupies a small portion in nowadays power systems, even in the system of countries that have developed wind generation for a long time, the proportion of wind energy is no more than 10 percents. Wind generator usually needs to absorb reactive power form system to maintain its operation, which makes wind generator seem like power user in the system. To keep a reasonable voltage of wind generator might be more necessary than to optimize system operation with considering this voltage as a variable. From another viewpoint of wind generator, the operation of wind generating is mainly to trace the variation of nature wind conditions, which is quite different from conventional generation that mainly operates to serve the users' fluctuating demands. As a result, it is hard to adjust wind generating to meet the system optimization requests. If take terminal voltage of wind generator as a constant value in optimization, the calculation can avoid heavy computational load and the result is assuringly a precise solution. In other words, the wind generator node can be taken as a PQ node with constant voltage in the optimizing process. A new optimization model based on goal programming is proposed
from this viewpoint. In this model, there are two objectives to optimize, the first one is to keep the voltage of wind generation at a given constant, and the second is to minimize grid active losses just as before. The model is described below: lexmin s.t.
g(u, x) 0 h(u,x)< O
=
2d
(6
from the target levels. In the real-world situation, the goals are achievable only at the expense of other goals and these goals are usually incompatible. It is necessary to establish a hierarchy of importance among these incompatible goals so as to satisfy as many goals as possible in the order specified. The approaches of nonlinear GP are summarized by Saber and Ravindran 131. The efficiency of these approaches varies with different problems. Among these methods, genetic algorithm can cope with complex nonlinear GP but have to spend more CPU time.
where Vg is the goal value of voltage control, xi denotes the IV. SIMULATION AND ANALYSIS terminal voltage of wind generator bus, di is the negative Based on the preparations in part II and part III, the deviation, PI,,, is the given expected value of grid active losses, simulation is carried out in a modified IEEE 14-bus system f(u,x) is the losses calculated, d2+ is the positive deviation. (Fig.2 ). In the following parts, the new model is called Constant Voltage PQ model (CVPQ) for short.
III. GENETIC ALGORITHM AND GOAL PROGRAMMING
A. Genetic algorithm
Genetic algorithm (GA) is a stochastic search method for 7 8 optimization problems. One of the important technical terms in GA is chromosome, which stands for a feasible solution of problems. GA contains five steps: initialization, evaluation, selection, crossover and mutation GA starts with an initial set of randomly generated chromosomes called a population. The number of individuals in the population is a predetermined integer that is called Fig. 2. The modified IEEE 14-bus system population size. All chromosomes are evaluated by the The synchronous generator at bus 8 is substituted for a so-called evaluation function. A new population will be wind farm consisting of fifty wind generators. The capacity of formed by selection process. The cycle from one population to single wind generator is 600kW, and other corresponding the next one is called a generation. After performing the parameters are list in appendix I. A constant shunt susceptance genetic system a number of cycles, the best chromosome is compensator at bus 8 is set as 0.15 p.u.. The rest data of buses regarded as the optimal solution of the optimization problem. and branches are the same as that of a standard IEEE 14-bus B. Goal programming system['4]. In optimal operation, bus voltage has upper limit 1.06p.u. GP(Goal Programming) was developed by Chames and Cooper[12]. GP can be regarded as a special compromise model and lower limit 0.94p.u.. The output limits for synchronous for multi-objective programming and has been applied in a generators power output are list in appendix II. In the optimizing model, the decision vector u consists of wide variety of real-world problems. In multi-objective decision-making problems, it can be the voltage magnitude of bus 1, 2, 3, 6 and the active output of assumed that the decision-maker is able to assign a target level generators at bus 2, 3, 6. Supposed the wind farm at its highest active output 30MW, for each goal and the key idea is to minimize the deviations
ORPF problem is solved with three optimizing models and the results of generator buses are list separately in Table I, II, III. In optimizing process of iteration model, grid active loss is reduced to 1.571MW, but the computing time is terribly long, nearly several times than that consumed by PQ model. In PQ model optimizing, grid active loss is 1.575MW and the grid state variables are very close to that in iteration model optimizing. So if the deviations are acceptable, PQ model is really a better choice. In CVPQ model optimizing, grid active loss is a little larger, but the voltage magnitude of bus 8 is restricted to 1.000p.u. exactly.
TABLE I
TABLE IV
OPTIMAL SOLUTION WITH DIFFERENT CONDITIONS
WG
Power
Iteration
PQ
Ploss
CVPQ
Ploss
1.571 1.804 2.111
V8m
0.993 1.004 1.010
V8m
0.989 0.999 1.009
Ploss
V8l
1.000 1.000 1.000
30MW
20MW
1OMW
Vm(pu)
1.050 1.050 1.046 1.060 0.993
TABLE II
Pg(MW)
10.96 58.14 99.96 61.50 30.00
Qg(Mvar)
1.68 21.64 23.14 5.46 -15.01
1
2
3 6 8
Bus
Vm(pu)
1.044 1.044 1.041 1.060 0.989
Pg(MW)
12.49 55.87 99.89 62.32 30.00
TABLE III
Qg(Mvar)
0.51 20.85 22.49 8.23 -15.00
1
2
3 6 8
operator is satisfied with the approximate optimal solution, ORPF in wind generation integrated system can be solved as conventional power system, and many efficient algorithms can act as well. But if the terminal voltage of wind generator is an important variable for wind generating operation that is expected to keep a constant value during optimization, the ORPF should be transferred to a multi-objective problem. The CVPQ model is a suitable method for solving the new case. Compared with iteration model and PQ model, grid active loss in CVPQ model optimizing is usually a little larger, but the voltage of wind farm is absolutely controllable. In other words, during the optimizing process, the wind generation system is isolated from grid influences, which can be controlled as needed. Even in a conventional power system, the CVPQ model may be also useful in some optimization cases. For example, if it is discovered that some bus voltage is expected to be a given constant to ensure the stability of power system, GP model might be right the method to use.
V. CONCLUSIONS
Wind generation has been gradually changing the power source structure, which urges us to develop methods and Bus Pg(MW) Qg(Mvar) Vm(pu) techniques to meet new demands. It is crucial to model wind 1.059 10.59 1.42 generator appropriately in solving ORPF in wind generation 2 1.060 61.40 23.29 integrated system. 3 1.058 100.00 26.72 The iteration model is surely an exact method in solving 1.060 58.60 0.29 ORPF, but it is not a practical method due to tremendous CPU 8 1.000 30.00 -15.03 time consumed. In contrast, the PQ model is just an Further simulations go along with changing active output approximate method. If the deviation from optimal solution is of wind farm and the result is listed in Table IV. According to acceptable, it is really a good substitute for iteration model. these simulations, it is found that the PQ model is a fine The CVPQ model is based upon different viewpoint of approximation of iteration model on ORPF problems, and the dealing with wind generator. It takes wind generator as a solution is usually in a small range around the optimal one. If special PQ node. The optimal solution is accurate but the grid
OPTIMAL SOLUTION WITH GP MODEL
1
active loss is usually a little larger than the PQ model for incompatibility of multiple objectives.
VI. APPENDIX
[7]
of wind farm connected to power system," Power System Technology, vol. 28(20), pp. 28-32, Oct. 2004.
[8]
Wu Yi-chun, Ding Ming, "Power flow analysis in electrical power
The electrical parameters of single wind generator are shown in appendix I. The upper and lower limits for generator power output are shown in appendix II.
APPENDIX I ELECTRICAL PARAMETERS OF SINGLE INDUCTION GENERATOR
[9]
Andres E Feijoo, Jose Cidras, "Modeling of wind farms in the load flow
analysis," IEEE Trans. On Power Systems, vol. 15(1), pp. 110-115, Feb.
2000. [10] Prabha Kundur, Power System Stability and Control fM]. McGraw-Hill
x1 (p.u.)
X2 (p.U.) r2 (p.u.)
xm (p.u.)
Rated slip
Power factor
APPENDIX II
Companies, 1994.
[11] Boading Liu, Theory and Practice of Uncertain Programming fM]. New
Bus 1 2
Pmax
Pmin 0 0 0 0
Qmax 10 50 40
24
3 6
[3]
B Venkatesh, G Sadasivam, M Abdullah, "A new optimal reactive power scheduling method for loss minimization and voltage stability
voltage control," IEEE Power Tech Conference, pp. 1-8, Jun. 2003.
[6] R He, Z H Bic, "Multi-objective optimization of reactive power flow in
balancing market," 39th International Universities Power Engineering Conference, vol. 2, pp. 1125-1129, Sept. 2004.