Research Of Fuzzy Neural Network Controller Optimized Algorithm Based On Genetic Simulated Annealing Algorithm

Since the intelligence age has come, fuzzy control and neural network have gradually become a hot research subject for scholars. In addition, with the improvement of the requirement for the degree of intelligence, fuzzy neural network based on these two technologies has gradually developed and perfected and has been employed in various areas especially in the aspect of controlling. It has brought out some exciting outcomes. In order to improve the function of fuzzy neural network controller, many optimized algorithms have also been used in this field with the hope of achieving better effect. This paper is a case study which aims to put genetic simulated annealing algorithm into fuzzy neural network controller and let it play its role.According to the study of foreign and domestic researches, this paper has carried out its work mainly in the following aspects:(1)Structure changed network model is concluded from the analysis of the former model of fuzzy neural network. The main characteristic of this network model is that there is a layer whose structure is changed. Through the adjustment of this layer, the structure of the network will be most feasible and simple. This paper has analyzed the features of genetic algorithm and simulated annealing algorithm, combining the two methods-genetic simulated annealing algorithm has perfected fuzzy neural network.(2)In the beginning of building the system of fuzzy neural network, it is needed to list all the rules because people do not know which rule is effective. In this case, the network will be complicated and finally it will affect the function of the whole system. Because of this, this paper has put forward an ideal strategy which contains two-step and two-stage in order to make the network structure simpler by the method of genetic simulated annealing algorithm. During the first step, genetic simulated annealing algorithm will adjust the network correspondingly. After this, there will be two stages for its optimization according to the result which has been achieved in the first step. First of all, in accordance with the rule matching degree of the previous piece, some ineffective rule nodes of previous piece will be deleted. Then further adjustment will be carried out according to the connection weight between previous piece and following piece. While the work of the second step is to further optimize the parameters in order to achieve the ideal effect of controlling under the background that the former structure will not be changed.From the result of the simulated test, the strategy with two-step and two-stage has ideally improved fuzzy neural network controller effectively through the way of using genetic simulated annealing algorithm and the control performance has been increased obviously.