| Summary: | 碩士 === 國立雲林科技大學 === 工業工程與管理研究所碩士班 === 100 === The more development of human civilization and economy, the more resource consumption, the development of solar and other substitute energy become an urgent issue. Diffusion process is a core processes in the solar cell. Its physical and chemical reactions and their corresponding product characteristics are non-linear, so engineers relying on experience cannot effectively amend the process. In this study, back-propagation neural network is used to construct a prediction module for diffusion process and genetic algorithm is combined to solve the optimization of process parameters. However, genetic algorithm has a main drawback: slow convergence. Therefore, this study proposes a revised multiple objective genetic algorithm (RMOGA) and an adaptive multiple objective genetic algorithm (AMOGA). The proposed methods use the concept of elite sets and local search. Besides, TOPSIS and Pareto sets for fitness of genetic algorithm are applied to solve the multi-objective problem. RMOGA and AMOGA enhance the breadth and depth of search and speed up convergence. Experimental results show that AMOGA has the best performance. Both of the quality and quantity of solutions of AMOGA are better than those of MOGA.
|
|---|
