结合LM算法的PID神经网络预测控制系统

A neural network-based predictive control system with levenberg-marquardt method

针对工业过程中非线性系统的复杂控制和实时性问题,提出1种基于LM算法的PID神经网络预测控制器的设计方法。该方法在原有的神经网络预测控制基础上,引入了1种新的神经网络结构与学习算法来建立预测模型,并针对该模型采用神经网络逆动态控制方法进行滚动优化以实现下一步控制量的优化。仿真结果表明,该方法有结构简单、计算量小、响应速度快等特点。在一定程度上满足复杂工业过程控制中的实时性要求。

Aiming at complicated control real-time problems of industrial nonlinear systems,a predictive control scheme based on PID neural network and Levenberg-Marquardt（LM）approach is proposed.On the basis of the existing neural network predictive control,this method introduces a new structure and learning algorithm of neural network,and a neural based dynamic inverse control algorithm is developed to realize the control optimization of the next time step for the model.Simulation results show that this method has simple structure,small computation burden and fast system response.To a certain extent,It satisfies the complex industrial process control real-time performance requirements.