基于RBF神经网络的控制器参数优化设计研究

以单输入单输出控制系统为研究对象,为快速完成控制器参数整定及优化,提升系统动态性能和稳态性能,提出了一种基于RBF(Radial Basis Function)神经网络的控制器参数优化方法。利用RBF神经网络的局部逼近能力和自学习能力,构造出控制系统辨识与控制器参数优化双网络结构,实现了对被控对象的在线辨识及增量式不完全微分PID控制器参数的在线迭代,快速完成控制器参数的整定,在保证系统动态特性的同时,大幅提升稳定精度。

Research on Optimal Design of Controller Parameters Based on RBF Neural Network

Taking single input single output control system as research object,a controller parameter optimization method based on RBF neural network is presented to quickly complete controller parameter tuning and optimization and to improve system dynamic performance and steady-state performance.Using the local approximation ability and self-learning ability of the RBF neural network,a control system identification and controller parameter optimization dual network structure is constructed.Taking this method,the online identification of the controlled object and the online iteration of the parameters of the incremental incomplete differential PID controller is realized.It can quickly complete parameter tuning,while ensuring the dynamic characteristics of the system,greatly improve the stability accuracy.