Some Improvements Of The Genetic Algorithm And Their Applications

During the research and practice of the scientific theory, there often exists such kind of problem that to determine the criterion of an optimal scheme, and the way of seeking the optimal scheme. This kind of problem is called an optimization problem. Scientific researchers have studied out rather mature algorithms and methods for some simple circumstance. However, for the complicated system, they usually feel incompetent and try to find more effective methods.Since 1940s, researchers have noticed the powerful adaptation ability of the living beings during the evolution of the nature. Superior genes are consistently duplicated, the adaptations of community are improved, and the excellent species with powerful adaptation emerges. The genetic algorithm is such optimal algorithm that is based on the research of nature adaptation theory. As a global optimization algorithm, the genetic algorithm has the following properties: (1) By the group search technique and implicit paralleling property, the efficiency of the algorithm is accelerated; (2) It relies only on the values instead of the algebraic expression of the objective function; (3) The reliability of the solution is improved because of the robustness of the algorithm. (4) It is simple and general, and easy to be combined with other algorithms to form hybrid algorithms.But the genetic algorithm is not efficient at local search. Although it can reach the nearby of the optimal solution quickly by the powerful global convergence property, it takes very long time to find the optimal solution. The main reason of the above phenomena is that the offspring constantly inherits the genes of the parent, which reduced the diversity of the community, so that the algorithm is trap into a local convergence. In order to make full use of the advantage of the algorithm, to avoid its weakness, and to speed up the convergence, one effective method is to adopt the hybrid strategy besides well designed basic Genetic Operators and properly chosen parameters of the algorithm. That is, to bring fast convergent local optimization algorithm into the genetic algorithm during the genetic evolution process, and to form a hybrid genetic algorithm.Foreign researchers have gained prominent results in enhancing the local search ability and speeding the optimization process of genetic algorithm (Allen Gardner B etal, 2003), (KurtA H etal, 2002), (Ghasemi M R etal, 1999). Domestic researchers also have gained many import achievements in various fields by studying the combination of the genetic algorithm with the flexible tolerance method or the simulated annealing method. The complex method (Zhan he Liu, 2004) was firstly proposed by J. Box in 1965, it optimizes the optimization individuals by the reflection, expansion, and compression operator. It posses powerful local optimization capacity, is easy to converge. This paper will study the combination of the genetic algorithm and the complex method, and analyze the performance of the hybrid algorithm through the case results.The paper contains six chapters. The first chapter introduces the character, evolution, and application of the genetic algorithm, and illustrates the paper's selected topic base, research contents and basic approach. The second chapter presents the basic knowledge of the genetic algorithm. The third chapter probes into the implementation technique of the genetic algorithm, introduces related basic genetic operators. The fourth chapter provides some improved strategies of the genetic algorithm. It not only improves the design of the basic operator and the selection of the algorithmic parameters, but also gives the improved scheme to solve the premature convergence phenomenon of the genetic algorithm. The improved scheme studies the hybrid of the genetic algorithm with the complex method from two points of the view, and is named the improved algorithm 1 and the improved algorithm 2. The fifth chapter applies the improved strategies of the last chapter to the distribution devices maintenance scheduling optimal model. Through the performance of the optimal results, the superiority of the improved genetic algorithm to the genetic algorithm is verified. The optimal results can reduce the loss to the large extent, and has powerful economic applied value. The sixth chapter gives the summary of the paper and the further research on the subject.