Composite neural networks adaptive control system of temperaturebased on GA learning
基于遗传算法学习的复合神经网络自适应温度控制系统

A kind of composite neural network adaptive control structure is proposed in this paper with the purpose of tackling problems such as nonlinearity and uncertainty in a temperature control system. The neural network positive model is constructed to represent the dynamic characteristics of the controlled object, and the neural network controller is used to realize the nonlinear mapping of optimal control rules. By taking eighty troops of history data to serve as samples the genetic algorithm with its searching ability and high efficiency is successfully used to train the weights of the multi-layer forward neural network. Then, the raising and falling temperature technics curve is acted as the input to simulate the temperature control system. The simulation results indicate the capability of genetic algorithm in fast learning of neural networks, guaranteeing a rapid global convergence and overcoming some shortcomings of the traditional error back propagation algorithms. It is shown that if this neural network adaptive control structure is applied to the temperature control system, the output of the neural network controller can adapt to changes of the object parameters and environment, and hence the temperature control system will have a nice learning and self-adaptive capability and lead to a good control result.