基于神经网络的时滞非线性系统的广义预测控制

Generalized Predictive Control of Time-Delay Nonlinear Systems Based on Neural Networks

针对一类时滞非线性被控对象,提出一种基于RBF神经网络的广义预测自校正控制方案,在广义预测控制中,采用RBF神经网络建立被控对象的多步预测模型,并不断修正预测输出,提高预测输出的精度。控制器则采用GPC隐式修正算法,不用辨识对象的模型参数,大大减少了计算量。经过仿真研究,与常规的PID自适应控制方法相比较,证明了该方法的优越性,预测控制误差小,实时性好,动态响应快。

A generalized predictive self-tuning control scheme based on RBF neural network is proposed for a class of time delay nonlinear controlled objects.In the generalized predictive control（GPC）,RBF neural network is used to establish multi-step predictive models of the controlled object,and constantly revising forecast output to improve the accuracy of predictive output.The controller adopts a GPC implicit correction algorithm,without to identify the model parameters,the calculated amount is gready reduced.By computer simulating,and comparing with the conventional PID adaptive control methods,the superiority of the method is proved,and it has small predictive control error,good real-time performance and fast dynamic response.