基于遗传算法的神经网络自适应控制器的研究  

STUDY ON NEURAL NETWORK ADAPTIVE CONTROLLER BASED ON GENETIC ALGORITHMS

作　　者：刘宝坤[1] 石红瑞 王慧[1] 赵丛 

机构地区：[1]天津大学自动化系

出　　处：《信息与控制》1997年第4期311-314,320,共5页Information and Control

基　　金：国家教委专项基金

摘　　要：提出了一种基于遗传算法的神经网络自适应控制方法．该方法是针对ＢＰ算法训练神经网络控制系统时收敛速度慢、动态特性不够理想等不足，用改进的遗传算法来优化神经网络辨识器与控制器的参数，以提高控制系统的性能．仿真实验表明该控制器对于非线性、时变、滞后等对象都具有很好的控制精度。This paper presents a kind of adaptive neural network(NN) control method based on genetic algorithms(GA). In order to overcome the defects such as slow convergent speed and unsatisfied dynamic character when backpropagation(BP) is used to train the NN control system, the method using a modified GA to train parameters of NN is proposed to improve the performance of the control system. Simulation shows that the controller has fine accuracy, robustness and dynamic character when it is used to control nonlinear, time variable and time delayed systems.

关 键 词：神经网络 遗传算法 自适应控制 自适应控制器 

分 类 号：TP273.2[自动化与计算机技术—检测技术与自动化装置]