An Intelligent Algorithm For Weight Optimization Of Generalized Fuzzy Systems

Fuzzy system through the information or data to obtain the language reflect the relation-ship between input and output of a complex system,although it does not depend on the precise mathematical model,but with logical reasoning,numerical calculation and nonlinear function ap-proximation ability,especially using it can put the IF-THEN fuzzy rules into nonlinear mapping of systematic procedures.At present,many scholars at home and abroad have done extensive research on various kinds of intelligent algorithms,and their achievements are numerous.For example,a hybrid learning algorithm that integrates BP algorithm into GA algorithm can realize self-learning of fuzzy system,thus achieving global optimization and fast search.For another example,firefly algorithm is proposed based on the luminous behavior and mutual attraction of fireflies in nature,then the target optimization can be realized according to the constant updating of the luminance and attraction of fireflies themselves.Then it is of great theoretical significance to combine general fuzzy system with intelligent algorithm.This paper mainly focuses on the above two issues,and the specific contents are as follows:The first part:The topic selection background and research status.The second part:Preliminary knowledge.We mainly introduce some related knowledges of Mamdani fuzzy system,T-S fuzzy system,polygonal fuzzy number and arithmetic operation.The third part:Based on triangular fuzzy,the product of reasoning machine and fuzzy method is introduced into the average solution center Mamdani fuzzy system model,especially fuzzy set values for the triangular fuzzy number,when the current domain space subdivision theory and the optimization parameter method design Mamdani fuzzy system,the validity of the proposed algorithm is demonstrated by statistical t-hypothesis test in a concrete example.The fourth part:In the hybrid fuzzy system composed of the Mamdani fuzzy system and the TS fuzzy system,an error function is established by the weight gradient vector,and the weight parameter of the gradient vector is optimized according to the BP algorithm and genetic algorithm,and the fitness function and decoding conversion are used to design.Weight parameter optimiza-tion algorithm.Finally,the goal of global optimization of the weight parameters of the hybrid fuzzy system is achieved.The fifth part:According to the ordered representation of the polygonal fuzzy number and its linear operation,the model of non-homogeneous linear polygonal T-S fuzzy system is introduced,and the expression of the adjusting parameters of the linear part of the system is given.Secondly,the TS firefly algorithm was designed to update the position of possible particles by using the fire-fly's own luminance and attraction formula,so as to realize the global optimization of all regulating parameters(particles).Finally,the numerical example shows that the algorithm can not only avoid falling into local minimum and precocious convergence problems,but also has the advantages of global convergence and high optimization precision.