Research On Genetic Algorithm Application In Distribution Network Optimal Planning

With the rapid development of economy, electric power demand increases quickly, and in the process many problems of distribution network are exposed, such as obsolete equipment, weak network structure, inadequate power supply of network capacity, poor quality and low reliability of power supply, high costs of operation and maintenance, national widespread distribution networks can not meet the needs of social development, and urgently need to construct and rebuild. Therefore it is necessary to carry out research on distribution network optimal planning.Firstly, the dissertation analyzes the current situation of China's distribution network, and expounds mathematics models of distribution netwok structure optimal planning and swithch optimal planning that can efficiently incorporate the economy and reliability of distribution network optimal planning, and collaboratethe the benefit of power supply company and customer, it is fit for distribution network optimal planning in electric power market.Secondly, the dissertation systematically discusses the theory of Genetic Algorithm and introduces the main modified measures towards the Genetic Algorithm's defects of premature convergence, low ability of local searching, no norms in encoding and failing to express constraint conditions of problems, especialy discusses the Partheno-genetic Algorithm and improves its selection operator, gene reorganization operator, gene mutation operator and convergence rule. Following are the specific improvements:1. Adopting the selection ways of the best save strategy and two two competition can avoid premature convergence and super numeral string easily caused by wheel disk gambling option when the difference of individual fitness is very large, and closed competition problem easily caused by blind searching when fitness is very near.2. Adopting the hyperbolic curve reorganization probability and exponent mutation probability increasing with the relative maintenance generation of the best individual can avoid destroy the better schemaes in population that have adverse effect on evolutionary calculation when the probability is too large, and the lower speed of generating new individual that makes difficult to achieve the best solution when the probability is too small.3. Adopting the smallest maintenance generation of the best individual as ruleconvergence can make the most of the knowledge accumulation of Partheno-genetic Algorithm in the generation process;this rule is more reasonable than the traditional largest Partheno-genetic Algorithm rule convergence.Finally, the dissertation introduces the realization process of modified Partheno-genetic Algorithm application in distribution network structure optimal planning, designs program using Visual C++ development platform and calculates the examples using Genetic Algorithm, Partheno-genetic Algorithm and modified Partheno-genetic Algorithm, the results indicate that the modified Partheno-genetic Algorithm is superior than Genetic Algorithm and Partheno-genetic Algorithm in calculate efficiency and global convergence.Furthermore, the modified Partheno-genetic Algorithm is applied in distribution network switch optimal planning and the example is calculated.The researches indicate that the modified Partheno-genetic Algorithm is efficiency and feasible application in distribution network optimal planning.