Fuzzy Neural Network Control System Based on Hybrid Learning Algorithm Liu Meijun Department of Electronic and Electrical Engineering, Xiamen University of Technology (Xiamen 361024) Abstract: Aiming at the uncertainty of complex nonlinear systems in the control process and the time-varying nature of parameters, a fuzzy neural adaptive predictive control system is designed to improve the accuracy of predictive control through error compensation; the learning algorithm of fuzzy neural network (FNN) is studied, and the global search ability of genetic algorithm is used to optimize the parameters of FNN controller offline, and the genetic operation is improved so that it can eventually search to the global optimum or the vicinity of the global optimum, and then the local search ability and adaptability of BP algorithm to the object are used to further adjust the parameters online. This makes the system have higher learning accuracy and faster convergence speed, and the obtained FNN has good generalization performance. The simulation results prove the effectiveness of this method. Keywords: fuzzy control neural network hybrid algorithm adaptive predictive control simulation A FUZZY NEURAL NETWORK CONTROL SYSTEM BASED ON HYBRID LEARNING ALGORITHMS LIU Mei-jun (Department.of Electronic and Electrical Engineering.,Xiamen University of Technology, Xiamen China 361024)ABSTRACT: For a class of complex nonlinear system with uncertainty and time-varying parameters in the process control,an adaptive predictive control system based on fuzzy neural network has been developed. using the error-compensation ,the accuracy of the system was improved; researching the algorithms of fuzzy neural network(FNN), an optimal or suboptimal spot is found by the optimization of fuzzy network ‚s parameters using the global searching ability of genetic algorithms.the BP algorithms ability of local searching and adaptation to object is used to adjust the parameters further. So the system owns more accurate precision and faster convergent speed ,and the FNN obtained has excellent performance of generalization.A simulation example demonstrates the efficiency of the method.KEY WORDS: Fuzzy control, Neural network, Hybrid algorithms, self-adaptive control. Simulation 1 Introduction In recent years, adaptive control of nonlinear dynamic systems has been a very active research field [1,2]. Fuzzy neural network control has been widely studied as an important adaptive scheme [3]. Fuzzy logic imitates the logical thinking of the human brain and is used to deal with systems with unknown or imprecise objects; neural networks imitate the functions of neurons in the human brain and can be used as general function estimators to map the relationship between system input and output. Fuzzy systems and neural networks are integrated with each other to construct various fuzzy neural networks as fuzzy information processing units to realize the automatic processing of fuzzy information. However, conventional fuzzy neural networks often have technical difficulties in the design process, and their control accuracy and learning ability need to be further improved. At the same time, the controller designed by the neural network is often not the global optimal, which is mainly due to the fact that the BP algorithm itself is easy to converge to the local minimum [4,5]. Genetic algorithm is a random search algorithm proposed by Professor J. Holland in 1975 that imitates the principle of biological evolution. Since its inception, it has been successfully applied in many fields such as function optimization, pattern recognition, image processing, and artificial intelligence. The main features of genetic algorithms are group exploration strategies and information exchange between individuals in the group. The search does not rely on gradient information. It has strong adaptability, robustness and global search capabilities, so it has been widely valued in the application of nonlinear function optimization problems. However, genetic algorithms also have the disadvantages of decreased search efficiency when approaching the optimal solution and possible premature convergence. In order to overcome the above shortcomings, this paper proposes a structured fuzzy neural network and designs an adaptive predictive control scheme. It adopts a hybrid learning algorithm that combines genetic algorithms and BP algorithms to achieve better control accuracy and nonlinear processing capabilities.
You Might Like
Recommended ContentMore
Open source project More
Popular Components
Searched by Users
Just Take a LookMore
Trending Downloads
Trending ArticlesMore