Sei sulla pagina 1di 5

Reactive Power Planning using Hybrid Evolutionary

Programming Method
V.Gopalakrishnan P.Thirunavukkarasu R.Prasanna
Lecturer in EIE Dept P.G.Scholar P.G.Scholar
gopalgct@hotmail.com thiruna_789@yahoo.com prasi@yahoo.com
GOVERNMENT COLLEGE OF TECHNOLOGY, COIMBATORE-13, TAMILNADU, SOUTH INDIA, INDIA.

Abstract: The paper proposes an application of Hybrid NQglim = set of numbers of buses at which reactive power
Evolutionary Programming to Reactive Power Planning. Reactive generations violate the limits.
power planning is a non-smooth and non-differentiable optimization
problem for a multi-objective function. The objective functions deals 1. INTRODUCTION
with the minimization of operating cost by reducing real power loss,
improving the voltage profile and minimizing the allocation cost of
additional reactive power sources. The proposed method is developed
In general, Reactive power planning (RPP) can be defined
in such a way that a standard evolutionary programming (EP) method as to determine the amount and location of shunt reactive
is acting as a base level search, which makes a quick decision to power compensation devices needed for minimum cost while
direct the search towards the optimal region, and local optimization keeping an adequate voltage profile. RPP is one of the most
by direct search and systematic reduction in size of search region complex problems of power systems, as it requires the
method is next employed to do fine tuning. The Reactive Power simultaneous minimization of two objective functions. The
Planning using Hybrid Evolutionary programming is demonstrated first objective function deals with the minimization of real
with the IEEE 30-bus system. The comprehensive simulation results power losses in reducing the operation cost and improving the
show that Hybrid Evolutionary Programming is a suitable method to voltage profile. The second objective function minimizes the
solve the Reactive power-planning problem.
allocation cost of the additional reactive power sources. RPP
Key words: - Evolutionary Programming, Hybrid Algorithms
Optimization, Reactive power planning.
is a non-linear optimization problem for a large-scale system
with many uncertainties. During the last decades there has
List of symbols been a growing concern about RPP problems. Conventional
Ni = set of numbers of load level duration calculus based optimizations algorithms have been used in
NE = set of branch numbers RPP for many years [1]. Most conventional optimization
Nc = set of numbers of possible VAR Source installment buses methods are based on successive liberalizations, which use the
Ni =set of numbers of buses adjacent to bus i, including bus i first and second derivatives of the objective functions and its
NPQ = set of numbers of PQ-buses, which are load buses with constraints equations as the search directions. Because the
constant P and Q injections formulae of the RPP problem are hyper quadric functions,
Ng = set of generator bus numbers such linear and quadratic treatments induce many local
NT = set of numbers of tap-setting transformer branches minima. Furthermore, the conventional methods cannot deal
NB = set of numbers of total buses with the non-differentiable factor in the reactive power
h = per unit energy cost ($/kWh) sources installation function in RPP. The conventional
dl = duration of load level (h) optimization methods can only lead to local minimum and
gk = conductance of branch k (p.u) sometimes result in divergences in solving in RPP problems.
Vi = voltage magnitude at bus i(p.u) Recently, new methods based on artificial intelligences have
Qij = voltage angle difference between bus i and bus j (rad) been used in RPP to solve local minimum problems and
ei = fixed reactive power source installation cost at bus i uncertainties [3, 5].
Cc I = per-unit VAR source purchase cost at bus i (f/p.u.VAR, EAs offer new tools for the optimization of complex
SB=100MVA) problem. Evolutionary algorithms have a number of
Qi = reactive power injected into network at bus i advantages over other conventional optimization and search
Qi = reactive power source installation at bus i (p.u) techniques. The EAs have attractive features such as the
PiQi = real and reactive powers, respectively, injected into simplicity of the algorithm, the ability to handle all sorts of
network at bus i (p.u) functional representations of problems, including very
GijBij= mutual conductance and susceptance, respectively complex function relationships. Despite their superior search
between bus i and bus j (p.u) ability, EAs fail to meet the high expectation that theory
NB-1= set of number of total buses excluding a slack bus predicts for the quality and efficiency of the solution. Genetic
GiiBii =self conductance and susceptance, respectively, of algorithms (GAs) have been given [3, 6] for the global optimal
bus i (p.u) solutions of reactive power optimization problems. Simulated
Qgi = reactive power generation at bus i annealing (SA) would be more likely to either prematurely
Tk = tap-setting transformer branch k converges or keep searching without direction GAs
NvPQlim = set of numbers of PQ-buses at which voltages consistently performs better than SA. Evolutionary
violate the limits programming robust and randomized algorithms, which can
adaptively search the globally optimal point. One
disadvantage of EP in problems is its slow convergence to a The objective function, therefore, can be expressed as
good near optimum. follows:
In general, Hybrid methods involve interwoven levels of
optimization. Namely evolution by any one of the min f C = I C + W C
evolutionary algorithms and individual learning by
conventional optimization methods, which cooperate in a
global process of optimization. s.t ,0 = Q i − V i ∑ V j ( G ij sin θ ij − B ij cos θ ij )iε N PQ
The paper proposes an application of a two-phase jε N i
evolutionary programming method based on the hybrid min max
method has been developed [9]. This method was also not able Q ci ≤ Q ci ≤ Q ci iε N c
to find the optimal solution for difficult constrained non-linear
min max
optimization problems consistently, because once again Q gi ≤ Q gi ≤ Q gi iε N g
modified evolutionary algorithm was employed in the phase-2
min max
of the algorithm. To get a compromise between the T k ≤ T k ≤ T k iε N T
computational expenses and the accuracy of the solution, a
new Hybrid EP method is proposed in this paper to improve min ≤ max iε
vi vi ≤ vi NB
the speed and solution quality so as to take challenging non-
(4)
linear constrained optimizations problems. The propose hybrid
Where reactive power flow equations are used as equality
techniques could outer from the conventional EP solution by
constrains. VAR source installment restrictions, reactive power
using EP as a base level search towards the optimal region
generation restrictions, Transformer tap-setting restrictions
with limited computing time. Subsequently, a local search
and bus voltage restrictions are used as inequality constraints.
method an optimization by direct search with systematic
Qcimin can be less than zero and if Qci is selected as a negative
reduction of the size of search region is employed to do the
value, say in the light load period, variable reactance should
fine-tuning to obtain an exact global optimum with light
be installed at bus i. The transformer tap-setting T, generator
computational expenses. The procedure handles either
bus voltages Vg and VAR source installments Qc are control
inequality constraints or equality constraints and the feasible
variables so they are self-restricted. The load bus voltages
region do not have to be convex. Gradient and other auxiliary
Vload and reactive power generations Qg are state variables,
information’s are not required.
which are restricted by adding them as the quadratic penalty
In order to prove the effectiveness of the proposed hybrid
terms to the objective function to form a penalty function.
algorithm, this method was applied to Reactive power
Equation (4) is therefore changed to the following generalized
planning. The IEEE 30-bus system is used in this paper as the
objective function.
simulation system. The results obtained were compared with
the results of Evolutionary programming methods. The
outcome of the study clearly demonstrates the effectiveness
min F c = f C + ∑ 2
λ vi(V i − V ilim ) +
iε N v lim
and robustness that a hybrid evolutionary programming (HEP)
can achieve over evolutionary algorithms. ∑ λ Qgi (Q lim 2
gi − Q gi )
iε N Qg lim
2. PROBLEM FORMULATION
s.t ,0 = Q i − V i ∑ (V j G ij sin θ ij − B ij cos θ ij )iε N PQ
The objective function in RPP problem comprises two jε N i
terms [6]. The first term represents the total cost of energy loss (5)
as follows: Where λVi and λQgi are the penalty factors, which can be
WC = h ∑ d l P loss , l increased in the optimization procedure; Vilim and Qgilim are
lε n l defined in the following equations:
(1)
Where Ploss,l is the network real power loss during the period ⎧ min if V p V min
lim = ⎪V i i i
of load level l. The Ploss,l can be expressed in the following Vi ⎨
equation in the duration dl: ⎪⎩V imax if V i f V imax
P loss = ∑ ( )
g k V i2 + V 2j − V i V j cos θ ij
⎧Q min if Q p Q min
kε n Ε ⎪ gi gi gi
lim
(2) Q gi = ⎨
⎪Q max if Q gi f Q max
The second term represents the cost of reactive power source ⎩ gi gi
installments, which has two components, namely, fixed (6)
installment cost and the purchase cost.
∑ ( )
It can be seen that the generalized objective function Fc is a
IC = e i + C ci Q ci non-linear and non-continuous function. Furthermore, it
iε n C contains of a lot of uncertainties because of uncertain loads
(3) and other factors.
3. EVOLUTIONARY PROGRAMMING q

EP is different from conventional optimization methods. It


Wi = ∑ Wt (9)
t =1
does not need to differentiate cost function and constraints. It
uses probability transition rules to select generations. Each ⎧ fr
⎪1if u l p
individual competes with other individuals in a combined Wt = ⎨ fr+ fi
population of the old generation and the mutated old ⎪0, Otherwise
generation. The competition results are valued using a ⎩
probabilistic rule. The winners of the same number as the (10)
individuals in the old generation, constitute the next Where q is the number of competition; fr is the fitness value
generation [8, 11]. The EP procedure is briefly listed as of the rth randomly selected competitor in the combined
follows: population; fi is the fitness value of pi, ui is randomly selected
from a uniform distribution set, U (0, 1). When all individuals
Initialization: The initial population of control variables is pi, i =1,2…2m, obtain their competition weights, they will be
selected randomly from a set of uniformly distributed control ranked in descending order of their corresponding value Wi
variables ranging over their upper and lower limits. The .The first m individuals are transcribed along with their
fitness score fi is obtained according to the objective function corresponding fitness values fi to be the basis of the next
and the environment. generation. The values of maximum, minimum and average
Statistics: The maximum fitness fmax minimum fitness fmin the fitness and sum of fitness of this generation are then calculated
sum of fitness εf, and average fitness favg of this generation in step 2.
calculated as follows: Convergence test: If the convergences condition is not met,
⎧ f i ≥ f ∀ f , j = 1,...m⎫ the mutation and the competition process will run again. The
f max = ⎨ f ⎬
⎩ i j j ⎭
maximum generation number can be used for convergence
condition. Other criteria, such as the ratio of the average and
⎧ f ≤ f ∀ f , j = 1,...m⎫
f min = ⎨ i
the maximum fitness of the population is computed and
⎬ generations are repeated until
⎩ fi j j ⎭ f avg
m ≥δ
fε = ∑ fi f max
i =1 (11)
fε Where δ should be very close to 1, which represents the
f avg = degree of satisfaction. If the convergence has reached a given
m
accuracy, an optimal solution has been found for an
(7)
optimization problem.
Mutation: Each selected parent, for example Pi, is mutated
and added to its Population following the equation. 4. HYBRID EVOLUTIONARY PROGRAMMING

Pi + m, j = Pi , j + N ⎜
⎜ (
0, β x j max − x j min ) fi ⎞
⎟, j = 1, 2,....n
f max ⎟
In general, local search techniques have the advantage of
solving the problem quickly, though their results are very
⎝ ⎠
much dependent on the initial starting point; therefore they
(8) can be easily trapped in a local optimum. On the other hand in
Where n is the number of decision variables in an individual, evolutionary programming samples a large search space,
Pi,j denotes the jth element of the jth individual: N ( µ,σ2 ) climbs many peaks in parallel, and is likely to lead the search
represents a Gaussian random variable with mean µ and towards the most promising area. However, an EP faces
variance σ2. fmax is the maximum fitness value of the old difficulties in a fine-tuning of local search: it spends most of
generation, which is obtained in statistics. xjmax is maximum the time competing between different hills, rather than
fitness value of the old generation which is obtained in fitness improving the solution along a single hill that the optimal
value of the old generation which is obtained in statistics. xjmax point locates. Hence, if one makes use of the advantages of
and xjmin are the maximum and minimum limits of the jth both the local and EP techniques, one can improve the search
element; and β is the mutation scale, 0<β≤1, that could be algorithm both effectively and efficiently.
adaptively decreased during generations. If any mutated The proposed hybrid EP combines a standard EP and the
value exceeds its limit, it will be given the limit value. The phase-2 of the conventional search technique. EP takes the
Mutation process (1) allows an individual with larger fitness place of the phase-1 of the search, providing the potential near
to produce more offspring for the next generation. optimum solution, and the phase-2 of a search technique using
Competition: Several individuals (k), which have the best optimization by direct search and systematic reduction of the
fitness, are kept as the parents for the next generation. Other size of search region. Phase-2 algorithm is applied to rapidly
individuals in the combined population of size (2m-k) have to generate a precise solution under the assumption that the
compete with each other to get their chances for the next evolutionary search has generated a solution near the global
generation. A weight value Wi of the ith individual is calculated optimum.
by the following competition as follows:
4.1 Phase-1 Algorithm 5. NUMERICAL RESULTS
Reactive power planning using Evolutionary Programming
method is solved [11]. In this section, IEEE 30-bus system has been used to show
the effectiveness of the algorithm. The network consists of 6
4.2 Phase-2 Algorithm generator-buses, 21 load buses and 43 branches, of which 4
branches, (6, 9), (6, 10), (4, 12) and (28, 27) are under load
After the phase-1 is halted, satisfying the halting condition tap-setting transformer branches. The parameters and variable
described in the previous section, optimization by direct
limits are listed in table 1.The possible VAR source
search and systematic reduction of the size of search region
installment buses are Buses 6, 17, 18 and 27. All power and
method is next employed in the phase-2. In the light of the
voltage quantities are per-unit values and the base power is
solution accuracy, the success rate, and the computation time,
used to compute the energy cost.
the best vector obtained from the phase-1is used as an initial
Two cases have been studied. Case 1 is of light loads
point for phase-2. whose loads are the same as [10 ].case 2 is of heavy load
This direct search optimization procedure (phase-2 whose loads are twice as those of case1.one year energy loss
algorithm) is implemented as follows [9]: cost is used to access the possibility of installing the VAR
sources, so the duration of load level is given as in table 4.
Step 1: The best solution vector obtained from the phase-I of
the hybrid algorithm is used as an initial point Q (0) for phase- 5.1 Initial Condition: All transformer taps are set to 1.0. The
2 and an initial range vector is defined as initial generator voltages are set to 1.0.The generator bus
R (0) = RMF * Range (12) voltages are set in different values to share the loads evenly
Where range is defined as the difference between the upper among the generators as table 2. The loads are given as:
and lower bound of p and RMF is the range multiplication Pload = 2.834 Qload =1.262.
factor. The value of RMF various between 0 and 1. The Transformer tap-setting and minimum cost is obtained
Step 2: Generate Ns trial solution vectors Around P (0) using as in Table 5 and Table 6.
following relationship, Table 1 Parameters and Limits

P i = P(0 ) + R(0 ) * rand (1, n )


SB (MVA) h ($/Kwh) ei($) Cci($/kVAR)
(13) 100 0.06 1000 30
th
Where Pi is the i trial solution vector, * represents element-
by-element multiplication operation and rand (1, n) is a Table 2 Reactive Power Generative Limits
Bus 1 2 5 8 11 13
random vector, whose element values various from –0.5 to Qgmax 0.596 0.480 0.6 0.53 0.15 0.155
0.5. Qgmin -0.298 -0.24 -0.3 -0.265 -0.075 -0.078
Step 3: Test for the feasibility condition and calculate an
objective function value for each feasible trial solution vector. Table 3 Voltage and tap-setting Limits
Step 4: Find the trial solution set, which minimize the Vgmax Vgmin Vloadmax Vloadmin Tkmax Tkmin
1.10 0.90 1.05 0.95 1.05 0.95
objective functions and equate it to P (0).
P (0 ) = P best (14) Table 4 Duration of Load Level
Case 1 Case 2
Where Pbest is the trial solution set with minimum f (P). dl(hour) 8760 8760
Step 5: Reduce the range by an amount given by
R (0 ) = R (0 ) * (1 − β ) (15) 5.2 Optimal Results
Where β is the range reduction factor whose typical value is
Table 5 Transformer tap-settings
0.05. Branch (6,9) (6,10) (4,12) (28,27)
Step 6: The algorithm proceeds to step 2, unless the best Hybrid 0.978 0.997 1.092 0.903
solution does not change for a pre specified interval of EP 0.981 1.042 1.029 1.037
generations Ng2 or maximum number of iteration is reached Table 6 Minimum Cost
specifically, the phase 2 of algorithm stops if the termination WC($) FC($)
Hybrid 534,421 2,637,938
condition of given equation is satisfied. For the best feasible EP 538,740 2,698,500
solution vector at iteration t, Pb, i and generation t-1 .Pb, t-1.
P b, t − P b, t − 1 / P b, t ≤ p 2 (16) 5.3 Output Graph

for a sufficiently small positive value p2, and for all j, for
successive Ng2 (integer number) iterations.
The main reason for the success of phase-2 algorithm is
that lies in its local search ability. Since the values for the
variables for always chosen around the best point determined
in the previous iteration, there is a more likelihood of
convergence to the optimum solution.
6. CONCLUSION
P.THIRUNAVUKKARASU: was born in Thirumayam, Tamilnadu, India
The use of Hybrid Evolutionary Programming for the in August 1979.He presently doing P.G in Government College of
Reactive power planning of power systems has been reported. Technology, Coimbatore, India. His specialization is in Power system
Reactive power planning is an optimization problem of a non- Engineering.
linear, non-smooth and non-continuous function. This type of
R.PRASANNA: was born in Madurai, Tamilnadu, India in 1980.He
less well behaved function is met with in most engineering presently doing P.G in Government College of Technology, Coimbatore,
problems, so the devising, testing and refining of new India. His specialization is in Power system Engineering.
techniques for finding optimal solutions has become more
important with the advent of even more powerful computers.
The Hybrid Evolutionary Programming approach has been
evaluated on the IEEE 30-Bus power network. The
simulations show that Hybrid Evolutionary Programming
always leads to satisfactory results for multi- objective
Reactive power planning, especially in non-continuous and
non-smooth situations. The comparison shows that the Hybrid
evolutionary Programming method is more powerful for
global optimization problems of the non-smooth, non-
continuous function. To prove the effectiveness of the
proposed hybrid algorithm, it was applied to reactive power
planning. On comparison of the results it was observed that
the hybrid evolutionary programming was more effective than
the evolutionary programme method.

7. REFERENCES

[1]. A. Kishore and E.F.Hill, “Static Optimization of reactive power sources


by use of sensitivity parameters”, IEEE Trans. Power Apparatus and
systems, vol. PAS-90, No., pp.1166-1173, 1971.
[2]. S.Sachdeva and R.Billinton, “Optimum network VAR planning by
nonlinear Programming”, IEEE Trans.Power Apparatus and Systems, Vol
PAS-92 No., pp1217-1225, 1973.
[3]. S.Swarup, M.Yoshimi and Y.Izui, “Genetic algorithm approach to
reactive power planning in power systems”, Proceedings of the 5th Annual
conference of power and Energy Socity IEE japan, pp-119-124, 1994.
[4].K.H.Abdual-Rahman and S.M. Shahidehpour, “Application of fuzzy
sets to optimal reactive power planning with security constrains”, IEEE
Trans. Power systems, vol PWRS-9, No-2, pp 589-597, May 1994.
[5]. Y.T.Hsiao.C.C.Liu, H.D.Chiang and Y.L.Chen, “A new approach for
optimal VAR sources Planning in large scale electric power system”,
IEEE Trans. Power systems, Vol.PWRS-8.No.3, pp.988-996, Aug.1993.
[6]. L.L.Lai, Senior Member and J.T.Ma, “Application of evolutionary
programming to reactive power planning – Comparison with Nonlinear
Programming Approach”, IEEE Transaction on power systems, Vol 12, No
1, Feb 1997.
[7]. L.J.Fogel, “The Feature of Evolutionary Programming”, Proceeding of
the 24th Asilomar Conferences on Signals, Systems, and computers pacific
Grove, CA, 1990.
[8]. Kwang.Y.Lee, Senior Member, IEEE, and Frank F.Yang, “Optimal
Reactive power planning using Evolutionary Algorithm: A comparative
study for Evolutionary Programming, Evolutionary Strategy, Genetic
Algorithm, and Linear Programming”, IEEE Transaction on power systems,
Vol. 13, No. 1, Feb 1998.
[9]. Dr S Baskar, Dr P. Subbaraj and Dr M V C Rao “Hybrid Evolutionary
Programming Solution to short Term Hydro Thermal Scheduling Problem”,
IE Journal, Vol 82, March 2002.
[10]. K.Y.Lee, Y.M. Park and J.L. Ortiz, “A United approach to optimal
real and reactive power Dispatch”, IEEE Trans Power apparatus and
Systems, Vol. PAS-104,No 5,May 1985.
[11]. V.Gopalakrishnan and P.Thirunavukkarasu, ”Reactive Power Planning
using Evolutionary Programming Method”, “Proceedings of National
conference of Power Electronics in Energy Conservation” Feb 2004.

8. BIOGRAPHIES
V.GOLAKRISHNAN: was born in Tamilnadu, India. in May 1968. He
presently is working as a lecturer in the department of Electronic and
Instrumentation in government college of Technology, Coimbatore, India. His
area of interest is Power system protection, computer network security.

Potrebbero piacerti anche