基于遗传算法的神经网络自适应控制器的研究

Study of Neural Network Adaptive Controller Based on Genetic Algorithms

提出了一种基于遗传算法的神经网络自适应控制器的设计方法。该控制器主要由2部分组成,第一部分利用遗传算法搜索出一组神经网络控制器的初始权值,第二部分用历史输入输出数据训练神经网络自适应控制器,经过训练后其达到稳定的控制效果。通过Matlab仿真表明,该控制策略克服了BP算法对初始值敏感的问题,有效提高神经网络的控制效果,和常规PID控制器相比较,超调量小、抗干扰性强和鲁棒性好。

A neural network adaptive controller based on genetic algorithm is presented.It consists of two parts.In the first part,the weights of neural network controller are obtained by means of the genetic algorithm;in the second part,after trained by the input and output data,the neural network adaptive controller gains the stable results.The simulation results indicate that Matlab simulation improves effeciency of neural network and has robustness and anti jamming,meanwhile proves the superiority of this controller to the PID controller.