Professional Documents
Culture Documents
08 Chapter 3
08 Chapter 3
08 Chapter 3
CHAPTER 3
3.1 INTRODUCTION
Genetic Algorithm was used to solve the load flow problem (Yin
and Germay 1991 and Wong and Li 1996). However simple GA fails to
converge for large scale system. Modifications on GA were introduced to
obtain better performance (Wong and Li 1997 and Acharjee and Goswami
2008a and 2008b). PSO is a computational intelligence-based technique that is
not largely affected by the size and nonlinearity of the problem, and can
converge to the optimal solution in many problems (del Valle et al 2008). PSO
based algorithms were applied for power flow analysis (EL-Dib et al 2004 and
Acharjee and Goswami 2009). Even though these algorithms converge
quickly, they will not be reliable since PSO gets stuck in local optima.
The load flow equations are the power balance equations at each
bus, both active and reactive powers. The bus active and reactive powers are
specified at load buses (PQ buses) and active power alone is specified at each
PV bus. The powers can also be calculated using bus voltages. The load flow
solutions are the bus voltages which make the difference between the
specified and calculated powers as zero or within the tolerance (<0.001).
Therefore, the objective of the load flow is to find the voltage magnitudes and
angles of the different system buses that minimize the difference between the
specified power and the calculated power. Hence, the power flow problem can
be formulated as an optimization problem.
n
Pi ( Vi Vj Yij cos( ij j i )) Pgi Pdi (3.1)
j 1
n
Qi ( Vi Vj Yij sin( ij j i )) Pgi Pdi (3.2)
j 1
j=1:n
35
Minimize f( V , )
The objective of the load flow is to minimize ‘f’ to get the voltage
magnitudes and angles of the buses to satisfy the power balance equations.
x 'G
i x Gr1 Fs (x rG2 G
x r3 ) [ VjiD ; D
ji ]
5. PSO operator
i) Choose XG ([|Vji| ji]) as present particle.
ii) Generate pbest. For ith particle, if fitness1 is lesser than
fitness2, then pbest is selected from parent voltage, else
it is selected from target voltage.
pbest i x iG [ VjiG ; G
ji ]; if f (x iG ) < f (x 'G
i )
x "G
i [ VjiPSO ; PSO
ji ];
37
equations.
6. By comparing fitness1, 2 and 3, select the best particle for next
iteration
if f (x"G
i ) f (x iG ) f (x i'G ) then x Gi 1
x Gi [ VjiG ; G
ji ];
else if f (x iG ) f (x 'G
i ) f (x "G
i ) then x iG 1
x i'G [ VjiD ; D
ji ];
else x Gi 1
x "G
i [ VjiPSO ; PSO
ji ];
For all the case studies the PSO and DE parameters are set as
follows:
First the algorithms are tested on a simple 6 bus system (Wood and
Wollenberg 1996). Line data and bus data for this sample system are given in
Appendix 1. The power flow problem is solved using DE, MHPSO and
39
MDEPSO algorithms. The population size taken for this system is 50. The
simulation results are taken after 500 iterations. 20 independent trials are
considered for each algorithm. The best results are considered as the solution
of the particular algorithm. Table 3.1 shows the best solution obtained for bus
voltages of sample 6 bus system. From Table 3.1, it is clear that the results
obtained by all the algorithms are same as that of NR method. Table 3.2
compares the accuracy that is the power mismatch obtained by MDEPSO and
other algorithms. Statistical performances of these evolutionary algorithms are
given in Table 3.3. All the algorithms provide very low value of standard
deviation. This shows the consistency of these algorithms. Mean value shows
that all trials are able to provide the converged solution. Figure 3.1 shows the
convergence characteristics during the initial iterations (upto 40 iterations).
Figure 3.2 shows the same upto 300 iterations. If the maximum power
mismatch (less than 0.001) is set as the stopping criteria, then MHPSO
converges quickly. If the program is allowed to run upto 500 iterations, the
accuracy of MDEPSO is greater than DE and MHPSO. If the tolerance is set
as 0.001, MHPSO algorithm is preferred as it converges quickly. MDEPSO
takes more time for convergence. When the accuracy is considered as the
main criteria, MDEPSO is preferable since it gives very high accuracy.
Table 3.1 Bus Voltages of 6 bus system for power flow problem in polar
coordinate
Table 3.2 Comparison of power mismatches for 6 bus system for power
flow problem in polar coordinate
Table 3.3 Statistical performances of EAs for 6 bus system for power
flow problem in polar coordinate
Table 3.4 Demand Data of sample 6 bus system with increased Real
Power Demand at 4th bus
Table 3.5 Bus voltages of 6 bus heavily loaded system for power flow
problem in polar coordinate
Figure 3.3 Convergence characteristics for 6 bus system for power flow
problem in polar coordinate (Heavily loaded system)
performance during the initial iterations and there is no change in the search in
later iterations. The bus voltage magnitudes obtained by using MDEPSO, DE
and MHPSO algorithms are represented as bar chart and are given in
Figures 3.5, 3.6 and 3.7 respectively.
Table 3.6 Statistical performances of EAs for IEEE 30 bus system for
power flow problem in polar coordinate
1.4
1.2
Magnitude of bus voltages
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25 30 35
Bus number
1.4
1.2
1
Magnitude of bus voltage
0.8
0.6
0.4
0.2
0
0 5 10 15 20 25 30 35
Bus number
To check the ability of the EAs to solve higher order systems, these
algorithms are tested on IEEE 118 system. The data are given in Appendix 3.
The population size is increased to 600. The performances are carried out for
20 trials and for each trial 500 iterations are considered. Statistical
performances of the evolutionary algorithms along with the time per iteration
are given in Table 3.7. MHPSO and DE can not solve the problem. Even
though MDEPSO is able to converge, it takes more time. High value of
standard deviation shows that, there is a large difference in the results during
the trials. High mean value indicates that only a few trials are able to give the
converged results. Best values show that DE and MHPSO are not able to
converge within the tolerance. The characteristic curves are given in Figures
3.8. Even though, MDEPSO takes more time for single iteration, it converges
within 100th iteration and so the total computational time is lesser. MHPSO
settles to the final value quickly but it doesn’t give the accurate result. Bus
47
Table 3.7 Statistical performances of EAs for IEEE 118 bus system for
power flow problem in polar coordinate
Figure 3.8 Convergence characteristics for IEEE 118 bus system for
power flow problem in polar coordinate
48
1.4
1.2
0.8
0.6
0.4
0.2
0
5 10 15 20 25 30
bus number
Figure 3.9(a) Magnitude of bus voltages for IEEE 118 bus system using
MDEPSO (from buses 1 to 30)
1.4
1.2
bus voltage magnitude (p.u)
0.8
0.6
0.4
0.2
0
35 40 45 50 55 60
bus number
Figure 3.9(b) Magnitude of bus voltages for IEEE 118 bus system using
MDEPSO (from buses 31 to 60)
49
1.4
1.2
bus voltage magnitude (p.u)
0.8
0.6
0.4
0.2
0
65 70 75 80 85 90 95 100 105 110 115
bus number
Figure 3.9(c) Magnitude of bus voltages for IEEE 118 bus system using
MDEPSO (from buses 61 to 118)
Pi Ei j
(G ijE j BijFj ) Fi j
(G ijFj BijE j ) Pgi Pdi (3.4)
Qi Fi j
(G ijE j Bij Fj ) Ei j
(G ijFj BijE j ) Q gi Qdi (3.5)
50
Minimize f (E, F)
f i
Pi2 Qi2 (3.6)
1. Initialize a set of random values for real (E) and imaginary (F)
parts bus voltages. Eji is generated between 0.9 p.u and 1 p.u
and Fji is generated from 0.0 p.u to -0.1 p.u. Random numbers
for E is generated for PV and PQ buses. Random numbers for
F is generated for only load or PQ buses. F for PV bus is
derived from specified voltage magnitude and the generated E.
xi = [Eji ,Fji]; i = 1 to Np; j = 1 to n except slack bus
Np -Population size n = number of buses.
x 'G
i x r1G Fs (x rG2 G
x r3 ) [E Dji ; FjiD ]
5. PSO operator
iv) Using Equations (2.1) and (2.2), update the velocity and
position of XG and get the new voltage vector.
x "iG [E PSO PSO
ji ; Fji ];
equations.
52
else if f (x iG ) f (x 'G
i ) f (x "G
i ) then x iG 1
x i'G [E Dji ; FjiD ];
else x Gi 1
x "G
i [E PSO PSO
ji ; Fji ];
Table 3.8 Real and Imaginary part of bus Voltages for 6 bus system
for power flow problem in rectangular coordinate
MDEPSO MHPSO DE
7.0749*10-6 9.2897*10-4 3.8574*10-5
54
Table 3.10 Statistical performances of EAs for 6 bus system for power
flow problem in rectangular coordinate
0.05
DE
0.045 MHPSO
MDEPSO
0.04
0.035
0.03
Power mismatch
0.025
0.02
0.015
0.01
0.005
0
0 0.5 1 1.5 2 2.5
function evaluation 4
x 10
Figure 3.10 Convergence characteristics for 6 bus system for power flow
problem in rectangular coordinate
The algorithms are also tested on large scale power systems. For
demonstration IEEE 30 bus system and IEEE 118 bus systems are considered.
The population size taken for these cases are 150 and 600 respectively. The
simulation results are taken after 500 iterations. The statistical measures like
mean, best, and standard deviation are calculated from 20 independent trials.
55
Table 3.11 Statistical performances of EAs for IEEE 30 bus system for
power flow problem in rectangular coordinate
Table 3.12 Real and Imaginary part of bus Voltages for IEEE 30 bus
system for power flow problem in rectangular coordinate
DEPSO DE HPSO
Bus
Real Imaginary Real Imaginary Real Imaginary
No
part part part part part part
1 1.06 0 1.06 0 1.06 0
2 1.0404 -0.0979 1.0404 -0.0979 1.0403 -0.0988
3 1.0122 -0.1338 1.0122 -0.1337 1.012 -0.1351
4 0.9988 -0.1632 0.9988 -0.1631 0.9985 -0.1648
5 0.9794 -0.2469 0.9794 -0.2467 0.979 -0.2481
6 0.9922 -0.1939 0.9923 -0.1937 0.9909 -0.1954
7 0.9779 -0.2231 0.9779 -0.2229 0.9766 -0.2244
8 0.9887 -0.2065 0.9887 -0.2063 0.9883 -0.2083
9 1.0193 -0.256 1.0194 -0.256 1.0185 -0.2593
10 1.0061 -0.2826 1.0061 -0.2826 1.0049 -0.2866
11 1.0494 -0.2636 1.0494 -0.2635 1.0486 -0.2669
12 1.0213 -0.2724 1.0213 -0.2723 1.0203 -0.2761
13 1.0348 -0.276 1.0348 -0.2759 1.0338 -0.2797
14 1.0035 -0.2844 1.0025 -0.2841 1.0014 -0.2881
15 0.9982 -0.2846 0.9982 -0.2846 0.9961 -0.2885
16 1.0069 -0.2795 1.0069 -0.2795 1.0049 -0.2832
17 1.0005 -0.284 1.0005 -0.284 0.9983 -0.2879
18 0.9855 -0.2925 0.9855 -0.2925 0.9843 -0.2965
19 0.9827 -0.2949 0.9827 -0.2949 0.9805 -0.2986
20 0.9828 -0.2927 0.9875 -0.2927 0.9854 -0.2963
21 0.9829 -0.287 0.9923 -0.287 0.9902 -0.2906
22 0.983 -0.287 0.9924 -0.2867 0.9913 -0.2906
23 0.9831 -0.2884 0.9857 -0.2883 0.9845 -0.2923
24 0.9832 -0.29 0.98 -0.29 0.978 -0.2933
25 0.9833 -0.2815 0.9783 -0.2815 0.9765 -0.2842
26 0.9834 -0.2836 0.959 -0.2835 0.9581 -0.2865
27 0.9835 -0.2742 0.9866 -0.2741 0.9849 -0.2765
28 0.9836 -0.2038 0.9862 -0.2037 0.9858 -0.2057
29 0.9837 -0.2895 0.9614 -0.2894 0.9596 -0.2918
30 0.9838 -0.3006 0.9454 -0.3006 0.9445 -0.3032
57
Table 3.13 Statistical performances of EAs for IEEE 118 bus system for
power flow problem in rectangular coordinate
Standard Number of
Algorithm Best Worst Mean
Deviation successful trials
Figure 3.12 Convergence characteristics for IEEE 118 bus system for
power flow problem in rectangular coordinate
Table 3.14 Real and imaginary part of bus voltages obtained using
MDEPSO algorithm for IEEE 118 bus system
3.4 CONCLUSION
The DE, MHPSO and MDEPSO algorithms are applied to solve the
power flow problem in polar and rectangular coordinates. For small scale,
lightly loaded system (6 bus), all the above algorithms work well. MHPSO
shows better convergence during the initial iterations. For heavily loaded and
higher order systems, the MDEPSO algorithm gives more accurate results
62