Optimal Reactive Power Dispatch Using Hybrid

Loop-Genetic Based Algorithm

Md Sajjad Alam Mala De
Student Assistant professor
Department of Electrical Engineering Department of Electrical Engineering
National Institute of Technology, Patna National Institute of Technology, Patna
Patna-800005, Bihar, India Patna-800005, Bihar, India
Email Id: sajjaddce22@gmail.com Email Id: mala@nitp.ac.in

Abstract: - Reactive power dispatch problem is one of the continuous and discrete variables. In the Last few years
important issues of the power system in the operation and different types of new methods are used for optimization
control. The Optimal Reactive Power Dispatch Problem is the problem such as evolutionary programming, genetic algorithm
minimization of active power loss in the power system by (GA) [2] particle swarm optimization (PSO), simulated
varying different system control variables, i.e. generator bus annealing [3], ant bee colony (ABC). GA is a part of
voltages or its reactive power output, transformer taps setting, evolutionary programming. In GA start with random
shunt Var compensator etc. The minimization of active power generated population and evolved to better solution using
loss will consequently reduce production cost for the genetic operators (selection, crossover, mutation). Now a days,
requirement of reactive power support. In this paper a new Evolutionary computation Method has been used in different
algorithm is presented which is motivated by a hybrid area of research of the power system.
approach called as hybrid Loop-Genetic based algorithm
In Evolutionary algorithm (EA) [4], population of solution
(HLGBA).The hybrid Loop-GA based algorithm (HLGBA)
is developed with large no of generation to find the optimal
uses the benefits from global search, i.e. the genetic algorithm
solution. Hence designing of optimization problem is more
with less evolution process (only for global search space) and
expensive. To reduce computational budget a new approaches
then from local search for refining the solution with the
i.e. Hybrid Loop-Genetic Based Algorithm proposed in this
limited computation and time. This algorithm solve ORPD
paper. To improve evaluation process of optimization the
problem more efficiently. This new algorithm verified on
important approaches in literature is the hybridization. In
standard IEEE_14 bus system to analyze the computation,
hybridization, both global search and local search optimization
time and efficiency of algorithm.
are combined to improve the efficiency of the proposed
algorithm. Local search technique is fast and effective but
Keywords- Optimal reactive power dispatch (ORPD); Genetic
depends on initial point. For good starting point global search
Algorithm (GA); Loop Optimization; hybrid Loop-genetic
is used. This hybrid approach is also called as mimetic
based algorithm (HLGBA).
algorithm [5].
In Hybrid loop-Genetic algorithm ,there are two stages
I. INTRODUCTION firstly uses a genetic algorithm we have to find global search
In Power System operation and control the two main short area with less no of iterations because use of GA is only for
term problem is the optimal reactive power dispatch and finding global search. After the genetic algorithm, in second
optimal active power dispatch. In the present work reactive stages, we use the loop algorithm with starting point taken
power dispatch problem has solved. The short term reactive from GA output.
power dispatch problem works on minimization of total active
power transmission losses and the voltage profile is also II. PROBLEM FORMULATION
improved with constraints. The operational constraints (state Optimal reactive power dispatch problem optimizes or
and control variable) contains reactive power source, bus minimizes the real power losses by setting the control
voltages, phase angle and transformer tap setting [1]. ORPD is variables generator voltages or its reactive power output of
used to determine the optimal setting of generator voltage or generator, transformer tap and shunt compensator like
its reactive power output, output of synchronous condenser, capacitor etc.
shunt capacitor bank, Transformer on Load Tap Changing
(OLTC) [1]. The conventional gradient based optimization Min ∑pkloss=∑ ( + −2 cosθ ) (1)
algorithm is used for decades. Gradient variable is hard to
solve because it has many assumptions in the objective
Where k= (i,j); i∈NB (Total no of buses)
function like analytical and differential properties [1]. The
gradient based optimization also unable to handle both

978-1-4799-5141-3/14/$31.00 ©2016 IEEE

 satisfied by load flow and inequality constraints equation (4)
to equation (7) satisfied by limit the variables.
∑pkloss = total active power losses in the power system
III. PROPOSED HYBRID LOOP-GENETIC BASED
Gk(i,j)=conductance of the branch k(in p.u) ALGORITHM

Vi, Vj = voltage (pu) of bus i and j respectively Several classical method and stochastic method are used to
solve reactive power dispatch problem, e.g. linear and non-
linear programming, Gradient non-linear quadratic
θ = load angle between bus i and j (rad)
programming etc. Now days due to advancement of EA,
genetic algorithm, PSO, differential evolution is used. The
NE= total no of branches intention of proposed algorithm is to refine and improve
efficiency of optimal reactive dispatch solution by taking the
Subjected to advantage (Fast & Effective, Refine the solution quickly) of
Local Search and removing the disadvantage not feasible with
Equality constraints:- bad initial point of local search by identifying good region for
search space i.e. give good initial point using Global Search. If
Active power flows balance equation at all buses we able to identify good region for search space its gives
always good initial point which is possible through Global
Pgi-Pdi-Vi∑ ( sin θij + Bij θij) = 0 (2) search. As discussed with good initial point local search is best
solution for ORPD. So if we combined global search and local
Reactive power flow balance equation at all buses search it will be best solution for ORPD with significant
computational cost.
Qgi-Qdi-Vi∑ ( sin θij + Bij θij) = 0 (3)
The algorithm based on the local search with global search.
Inequality constraints:- The two key approaches are used.

1. Global search :- The fist key approaches of the proposed

Reactive power generation limit for each bus
algorithm is global search using genetic algorithm with less
no of generation than the actual genetic algorithm because
≤ ≤ i∈Ng (4) GA find global search only. The output of GA is optimal
value of control variable with all constraints.
Voltage magnitude limit voltages for each bus 2. Local Search: - Now perform the local search using loop
optimization method. In this method for each control
≤ ≤ , i∈ NB (5) variable, loop are used with all constraints. Now
constraints variable limits are taken from global search
Transformer tap setting limits space i.e. output of GA.

≤ ≤ , i ∈ NT (6) A. Genetic Algorithm

Reactive power source installation Genetic algorithm is categorised as a global search

heuristic technique to find exact or approximate optimized
≤ ≤ , i ∈ Nc (7) solution. It is inspired from evolution process such as
selection crossover and mutation. The GA starts from a
Where Bij represents the mutual susceptance in bus i population of randomly generated individuals [6]. In each
and j. Pgi and Qgi represents the active power and reactive generation find the fitness value of each population and now
power generation of bus i. select the population for next according to the fitness value
after selection different genetic operator i.e. crossover,
Pdi and Qdi are the active power and reactive power demand of mutation etc., are used.
bus i. Vi is the voltages of the buses,Vimax and Vimin is the
upper and lower limits of the voltages. Ti transformer tap in Algorithm: Genetic Algorithm to Solve ORPD
branch k, Timax and Timin is the upper limit and the lower limit
1. Read the system input data (branch data, bus data from
of transformer tap setting. Qci represents the reactive power
IEEE-14 Bus system)
source installation and upper limit and lower limit value is Qci
max
and Qcimin. NB and Ng is the no of system buses and 2. Initialize the parameter for Genetic Algorithm population
generator buses respectively.NC and NT is the no of reactive size, Maximum no of generation, the probability of
source installation and no of transformer tap. crossover, and probability of mutation.
The equation (1) represents the total transmission loss in the 3. Set the control variable, i.e. Generator voltage and
system. Equality constraints equation (2) to equation (4) transformer tap and shunt capacitor bank.
4. Start generation(gen=0) 3. Now set the control variable (5 generator Bus) and for
each control variable for-loop are used.
5. gen=gen+1
4. For each and every increment of variable, run the load
6. population=0 flow (N-R) method.
7. population=population+1 5. Check if power flow is converged then find and store the
control variable and active power loss, go to 4.
8. Randomly generate the population of control variables
(generator voltage and transformer tap) within limits to 6. Set the active power loss with high value.
form a single string.
7. After terminating the loop, find the minimum active
9. For each population run the load flow using Newton power loss and corresponding control variable.
Raphson Method
8. Stop.
10. If optimal power flow is not converged, then go to step 8.
V. SIMULATION AND RESULT
11. Replace valid chromosome in the initial population.
The standard IEEE_14 Bus System is used to test the system
12. Calculate the fitness value (1/ (total active power loss)) to prove the accuracy and efficacy of the new algorithm for
from valid chromosome. Optimal Reactive Power Dispatch problem. The single line
diagram of IEEE_14 system is in fig 1. Description of IEEE
13. If the (population<=population size) then go to step 7. 14 Bus System is given in table 1 [9].Constraints of the
variables is given in table 2. Using MATLAB Code
14. Find the best fitness value from all the fitness value and
(MATLAB 2012a) hybrid Loop-Genetic based algorithm was
store the corresponding total active loss and control
simulated and then run on a PC. The results show the accuracy
variable (bus data and branch data).
of the proposed algorithm and application of the algorithm.
15. For producing the new population for next generation.
Select the parent for reproduction using roulette wheel It is simulated for two different case types -
selection.
Case 1: ORPD Problem solved by proposed method with 5
16. Perform five point crossovers from randomly selected control variable (5 Generator Buses).
parent.
Case 2: ORPD Problem solved by proposed method with 8
17. Perform the mutation from randomly selected parent. control variable (5 Generator Buses and 3 Transformer
Tap).
18. If (gen<=Max gen) go to step 5.
Firstly the total power loss for optimal load flow is calculated
19. Display minimum active power loss and control variable using N-R method. The total power losses found 13.4014 MW
[6] for the 14 bus IEEE standard test system.
20. Stop.

B. Loop Optimization

Loop optimization is a basic search optimization method for

solving problem which has good initial points i.e. it has good
search space. In this optimization method nested loop are used
to refine the solution after GA, for each variable a, for loop are
used with very small increment. This optimization search is
like pattern search so it is called as local search optimization
method [7].

Algorithm: Solve ORPD Using Loop Optimization

1. Read the system input data(branch data, bus data from

IEEE-14 Bus system) Control variable limits taken from
Global Search Space i.e. GA.

2. Initialize the parameter for Loop optimization method and

the initial value is taken from GA output.
 INPUT PARAMETER CONSTRAINTS The optimal values of control variables (seen in dark),
(Generator Voltage) (Initial Limits taken from GA) active power loses and reactive power loses obtained from GA
V1 1.070 – 1.090 and result of HLGBA are presented in Table 5.
V2 1.020 – 1.055
V3 1.000 – 1.500 TABLE 4. CONSTRAINTS( 5 CONTROL VARIABLES) FOR
LOOP OPTIMIZATION METHOD
V6 1.020 – 1.050
V8 1.020 – 1.050
S. No. Description Values
TABLE 5. RESULT OF ORPD USING GA AND HLGBA FOR 5
1 Buses 14 CONTROL VARIABLE
2 Branches 20
Parameter O/P Using O/P Using
3 Generators 5
GA HLGBA
4 Generator or buses 9 V1 1.0645 1.0900
5 Tap-Changing 3 V2 1.0400 1.0550
transformers V3 1.0100 1.0200
V4 1.0073 1.0197
V5 1.0099 1.0238
No. Variables constraints UnitsLower Upper V6 1.0432 1.0500
Limits Limits BUS
V7 1.0351 1.0417
VOLTAGES
1 Voltage Load Bus Pu 0.95 1.05 V8 1.0500 1.0500
(Pu)
2 Voltage Generator Bus Pu 0.90 1.10 V9 1.0297 1.0373
3 Trans. Tap Setting Pu 0.10 1.00 V10 1.0245 1.0320
Fig: 1. IEEE 14 Bus Systems V11 1.0302 1.0373
V12 1.0281 1.0350
V13 1.0232 1.0302
V14 1.0084 1.0159
Active Power Loss(MW) 13.2198 13.1229
TABLE 1. DESCRIPTION OF BUS SYSTEMS Reactive Power Loss 53.9376 53.4699
(MVAR)
TABLE 2. CONSTRAINTS OF VARIABLES
The total active power loss of ORPD using GA is 13.2198
MW. The voltage profile is improved after GA which can be
Case 1: seen in fig 2. Proposed algorithm(HLGBA) is applied to
minimize the active power loss in the system it can varies the
In this case, the ORPD of IEEE 14 bus system is optimized control variable such as total active power losses decreases
using hybrid loop-GA method with 5 control variables (5 from 13.40 MW to 13.1229 MW, which is equal to % 2
Generator bus voltages). This problem solved by optimal saving and the voltage profile is also improved, it can be seen
setting of generator bus voltages using the proposed algorithm. in fig 3. The reactive power loss is also improved from
Genetic parameter values are given in the table 3. 56.4234(N-R Method) to 53.4699 MVAR.

TABLE 3. GA PARAMETERS FOR 5 CONTROL VARIABLE

Sl. No. Parameters Values

1 Length of Chromosome 25
2 No of Generation 50
3 Population Size 25
4 String Length 25
5 Crossover Probability 0.9
6 Mutation Probability 0.2
Fig 2: Using GA Voltage Profile of Bus Voltages
Due to randomness nature of GA, The program runs for 20 (5 control variables)
times to reduce the effect of randomness. After result of GA,
loop optimization method used to refine the solution whose
limits of control variables taken from GA (best value and
worst value taken as a lower and upper limit) and initial value
initialized by the best value of GA presented in Table 4.
 V4 1.0080 1.0177
V5 1.0145 1.0209
V6 1.0503 1.0413
BUS V7 1.0234 1.0287
VOLTAGES V8 1.0184 1.0242
(Pu) V9 1.0239 1.0266
V10 1.0209 1.0215
V11 1.0319 1.0277
V12 1.0343 1.0260
Fig 3: Using HLGBA Voltage Profile of Bus Voltages V13 1.0285 1.0211
(5 control variables) V14 1.0074 1.0056
Case 2: Active Power Loss 12.4645 12.4645
In this case, The ORPD problem is optimized by hybrid (MW)
loop-GA algorithm with 8 control variables (5 Generator Reactive Power Loss 51.5754 51.5754
(MVAR)
buses and 3 transformer tap). This problem solved by optimal
The total active power loss of ORPD using GA is 12.4645
setting of generator bus voltages and trans. tap using the
MW. The voltage profile is improved after GA which can be
proposed algorithm. Genetic parameter values are given in the
seen in fig 4. Proposed algorithm (HLGBA) is applied to
Table 6. Due to randomness nature of GA, The program runs
minimize the active power loss in the system it can varies the
for 20 times to reduce the effect of randomness [6]. After
control variable such as total active power losses 12.4645
result of GA, loop optimization method used to refine the
MW, which is same as GA output, but the voltage profile is
solution whose limits of control variables taken from GA
improved which can be seen in the fig 5 and the active power
output (worst and best value) and initial value initialized by
loss is decreases from 13.40 MW to 12.4645 MW Which is
the best value of GA presented in Table 7.
%7 saving of the active power loss. It can be seen in figure 5.
TABLE 6. GA PARAMETER FOR 8 CONTROL VARIABLE The reactive power loss is also improved from 56.4234(N-R
Sl. No. Parameters Values Method) to 51.5754 MVAR.
1 Length of Chromosome 46 The optimal values of control variables (seen in dark),
active power loses and reactive power loses obtained from GA
2 No of Generation 50
and result of HLGBA are presented in Table 8.
3 Population Size 25
4 String Length 46
5 Crossover probability 0.9
6 Mutation probability 0.2
TABLE 7. CONSTRAINTS (8 CONTROLVARIABLES) FOR
LOOP OPTIMIZATION METHOD
INPUT PARAMETER CONSTRAINTS
(Generator Voltage) (Initial Limits from GA)
V1 1.070-1.080
V2 1.050-1.060
V3 1.025-1.030
V6 1.040-1.050
Fig 4- Using GA Voltage Profile of Bus Voltages
V8 1.020-1.025
(8 control variables)
T1 0.600-0.700
T2 0.500-0.600
T3 0.900-1.000

TABLE 8. RESULT OF ORPD USING GA AND HLGBA FOR 8

CONTROL VARIABLES
Parameter O/P Using GA O/P Using
HLGBA
T1 0.4173 0.6787
Transformer T2 0.9690 0.5700
tap`
T3 0.9320 0.9291
V1 1.0894 1.0781
V2 1.0426 1.0600 Fig 5-Using HLGBA Voltage Profile of Bus Voltages
V3 1.0048 1.0268 (8 control variables)
 VI. CONCLUSION
The paper presented a new approach: Loop-GA based
algorithm to solve ORPD problem and it is tested on IEEE 14
bus system. The most important advantages of hybrid Loop-
genetic based algorithm over other modern heuristics are less
computational time fast convergence and more accurate than
other method. It can be easily coded on MATLAB 2012 a.
Using HLGBA the total active power loss found 13.1229 MW
(5 generator bus voltages as control variables) and 12.4645
MW(5 Gen. bus voltages and 3 trans. tap setting as control
variables) which is reducing %2 and % 7 of 13.40 MW value
in optimal power flow. Hence we can say that HLGBA
approach can be accepted to solve this kind of problem for
future studies.

REFERENCES

[1] B. Zhao, C.X. Guo, and Y.J. Cao, an Improved Particle

Swarm Optimization Algorithm for Optimal Reactive
Power Dispatch, Power Engineering Society General
Meeting, 2005. IEEE, 272 - 279 Vol. 1, 12-16 June 2005
[2] Dawood Talebi Khanmiri, Nasibeh Nasiri, and Taher
Abedinzadeh, An Optimal Reactive Power Dispatch
Using an Improved Genetic Algorithm, International
Journal of Computer and Electrical Engineering, Vol. 4,
No. 4, August 2012
[3] Y. L. Chen and Y. L. Ke, “Multi objective VAr planning
forlarge-scale power systems using projection based two
layer simulated annealing algorithms,” IEE Proc. On
Gener. Transm. Distrib, vol. 151,no. 4, pp. 555-560, July
2004.
[4] K. Lee and F. yang, “Optimal reactive power planning
using evolutionary algorithms: a comparative study for
evolutionary programming, evolutionary strategy, genetic
algorithm and linear programming,” IEEE transactions
On Power Systems, vol. 13, no. 1, pp. 101-108, 1998.
[5] Hemant Kumar Singh, Amitay Isaacs and Tapabrata Ray,
A Hybrid Surrogate Based algorithm (HSBA) to solve
Computationally Expensive Optimization Problems,
IEEE Congress on Evolutionary Computation (CEC)
2014.
[6] Dawood Talebi Khanmiri, Nasibeh Nasiri, and Taher
Abedinzadeh , Optimal Reactive Power Dispatch Using
an Improved Genetic Algorithm, International Journal of
Computer and Electrical Engineering, Vol. 4, No. 4,
August 2012
[7] https://en.wikipedia.org/wiki/Loop_optimization
[8] Q H Wu*, Y J Cao and J Y Wen ,Department of
Electrical Engineering and Electronics, Optimal reactive
power dispatch using an adaptive genetic algorithm, The
University of Liverpool, Liverpool L69 3GJ, UK
[9] Syamasree Biswas (Raha), Trina Som, Kamal Krishna
Mandal and Niladri Chakrabort, Cuckoo Search
Algorithm based Optimal Reactive Power Dispatch,
International Conference on Control, Instrumentation,
Energy & Communication, 2014.