Genetic Algorithm And Its Application For Optimization Design Of Special Transformer

Optimization mathematical model of special transformer electromagnetism parameters is a nonlinear programming problem with equality and inequality constraints, to meet some certain design objective. In this thesis, preprocessing of the optimization mathematical model, optimization method and software framework of software architecture are studied.The main works of the thesis are summarized as follows:1) The optimization mathematical model for special transformer electromagnetism parameters is prospected, and two kinds methods of the preprocessing process for the above model are analyzed. One is section and decision making method, which solves the question of a complicated problem by dividing it into several simple problems. Another is a series of standardized treatment methods for the model components (i.e. objective function, optimization variables and constraints). Test result shows that the proposed preprocessing methods can improve the performance of the algorithm, simplify the optimization process, and the best solution quality gained is enhanced.2) A new construction method of dynamic encoding based on code table is presented. The key thought lies on constructing a referring encoding table (i.e., code table) using knowledge, and put forward a kind of collection dynamic and static characteristic encoding method. The search efficiency and the quality ofoptimum solution can be enhanced, and the commonability of the code method has been strengthened on certain degree.3) A novel culture operator culture operator is proposed in genetic algorithm. The operator can preserve good characteristics genes in individuals at a higher probability, and genetic operator operation proceeds in the most promising direction. The individuals' fitness of population can be increased rapidly at the initial process of genetic operation, and the individuals' diversity in the population is maintained in the evolution procedure, which is particularly outstanding in the genetic operation's end process.4) A self-adaptation genetic algorithm based on knowledge (SAKGA) is proposed. The convergence of SAKGA is analyzed and proven. The optimization data show that SAKGA can produce performance improvement in execution time and accuracy, and it is potential to solve engineering optimization problems.5) A new software architecture framework is presented, and the software system of optimization design for special transformer is developed. The framework can separate electromagnetic parameters formulation and performance computation and simulation from the whole optimization process, guide the direction of optimization design by the peculiar adjustment mechanism in the system. Thus it can overcome non-linear coupling relation of parameter effectively in traditional software framework, which makes adjustment of parameter extremely difficult.