Conclusion
In this paper, we discussed and verified the effectiveness of the proposed methodology of optimal allocation of fixed and switched capacitors for unbalanced radial distribution feeders. In test case 1, it was proven experimentally, that AI-based Genetic Algorithm (GA) is superior than SQP in terms of the generated savings. It was also observed that although GA operations are based on stochastic processes, the output solution for several trials converges on almost the same amount of savings.

GA is domain independent. It works on coded structures of variables instead of the actual variables. The only information needed is the objective function. For that reason, we had come up to a formulation that considers most of the simplifying assumptions employed by previous researchers.

GA is well suited to unconstrained combinatorial optimization problems. Constraints can be incorporated by adding a penalty to the objective function for any constraint violations. We did not consider the lower voltage limit as a constraint since it is not the primary objective of the project. Moreover, increasing the voltage profile above the lower limit must be handled by voltage regulators and not by shunt capacitors since shunt capacitors are primary used for loss reduction. On the other hand, based on the generated solution of the GA method, no over-voltages occurred at all load levels after compensation. This is because the additional losses due to over-voltages act as an adaptive penalty function for the fitness value that guide the search process to the permissible solution. 

There are no defined rules on selecting GA parameters like population size and mutation probability. Thus, determination of GA parameters are user dependent.  There are, however, researchers that propose methods of selecting GA parameters that are close to optimal for a better GA performance.

GA is like an iterative process. For each generation, fitness of each individual is evaluated using objective function. Hence, GA runs slower than other optimization methods.

Most of distribution feeders feature a non-three-phase lateral feeders. Using sensitivity analysis, it was observed that these laterals have a high sensitivity factor. Test case 4 shows that savings significantly increased if single phase and two-phase lateral feeders are considered on the optimization process.

References

1. H. N. Ng, M. M. A. Salama, and A.Y. Chikhani, IEEE Transactions on Power Delivery, 15, no. 1, 387-392(2000)
2. D. Shirmohammadi and C. S. Cheng, IEEE Transactions on Power Systems, 10, no. 2, 671-679(1995)
3. S. Sundhararajan and A. Pahwa, IEEE Transactions on Power Systems, 9, no. 3,1499-1507(1994)
4. K. N. Miu, H. D. Chaiang and G. Darling, IEEE Transactions on Power Systems, 12, no. 3, 1160-1166(1997)
5. http://www.techsoftpl.com/matrix/matlite.htm. (May 2003).
6. R. D. Sarmiento, Undergraduate Student Project, Department of Electrical and Electronics Engineering, University of the Philippines, Diliman,(2002)
7. R.M. Altamirano, M.G. Dacayanan, R.P. Gatbonton, R.M. Olea, J.D. Villanueva, Undergraduate Student Project, Department of Electrical and Electronics Engineering, University of the Philippines, Diliman,(2001)
8. P. Attaviriyanupap and B. Eua-Arporn, Optimal Power Flow Using Genetic Algorithm, Electrical engineering Department, Chulalongkorn University, Bangkok. http://ee30-si.eng.hokudai.ac.jp/phatom/papers/OPF-GA.pdf (August 2003).
9. T. Back, Mutation Parameters, www.iop.org (July 2003).
10. L. Davis, “Handbook of Genetic Algorithms,” International Thompson Computer Press, New York(1991)
11. K. Figueredo, et. al., Intelligent Solutions to Energy Conservation Problems, http://www.ica.ele.puc/rio.br/publicacoes/download/cnf/0002.pdf
    (June 2003.)
12. J. H. W. Hsiao, Literature Review on Genetic Algorithm, http://homepages.inf.ed.ac.uk/s0231692/Resume/GeneticAlgorithmNew.pdf (July 2003).
13. G. Kendall, AI methods, http://www.cs.nott.ac.uk/~gxk/aim/notes/geneticalgorithms.doc (August 2003).
14. N. Khan, Population Sizing in Genetic and Evolutionary Algorithms, Illinois Genetic Algorithm Laboratory, Department of General Engineering University of Illinois at Urbana-Champaign,
    http://www-dvancedgec.ge.uiuc.edu/literature%20reviews/Ge493litreview.doc
    (August 2003).
15. K. Y. Lee, et al., IEEE Transactions on Power Systems, 15, no.3,
    916-922(2000)
16. M. G. Rojo, M. S. Thesis, Department of Electrical and Electronics Engineering, University of the Philippines, Diliman(2000)
17. G. Rzevski, “Artificial intelligence in Engineering: Past, Present and Future,” The open University, Milton Keynes, UK(1995)
18. W. M. Spears, Adapting Crossover in a Genetic Algorithm, Naval Research Laboratory, Washington, D.C., http://hive.cs.uwyo.edu/_wspears/papers/ (July 2003).
19. W. M. Spears, Adapting Crossover in Evolutionary Algorithm, http://hive.cs.uwyo.edu/~wspears/papers/ (July 2003).

Introduction
Problem Formulation and Solution
Implementation, Testing and Results
Testing and Results (continued1)
Testing and Results (continued2)
Conclusion and References