一种基于模糊遗传算法的隶属函数自学习方法研究

On a Method of MembershipFunctions SelfLearningBased on Fuzzy Genetic Algorithm

将传统遗传算法在模糊系统中进行了推广。研究了一种模糊遗传算法，给出了算法描述，一般性讨论，并在模糊控制器参数自适应整定问题中得到应用。

In this paper we propose a new fuzzy genetic algorithm to solve membershipfunctions selflearning problems of fuzzy controllers. The performance of a fuzzy controller depends to some extent on the choice of the membership functions. However, the adjustment of parameters of the membership functions is by trialanderror. So it is obvious that application of fuzzy controllers to a practical industrial production process is difficult. Some methods, such as neural networks and genetic algorithm, were proposed by other authors to perform selfadjustment of the parameters. Here we present a new algorithm to solve the problem of fuzzy controller′s membershipfunctions selflearning: the Fuzzy Genetic Algorithm (FGA). FGA is a kind of directed random search and optimization algorithm over all fuzzy subsets of some interval; here, it is ‘directed’ in a way different from traditional purely random search. In this paper, we describe the elementary fuzzy genetic algorithm and discuss the method and its application to parameter selflearning problems of adaptive fuzzy controllers. Simulation results show that FGA has good performance in time delay, accuracy, and stability.