基于神经网络的自适应控制器在CSTR中的应用研究

Application of Adaptive Controller Based on Neural Network for CSTR System

针对工业过程的特点和控制要求,提出一种基于多步预测的神经网络自适应控制算法。该控制器采用改进的RBF神经网络对过程进行建模,利用多步预测误差对神经网络控制器进行训练,从而实现控制器参数的在线自适应寻优。针对CSTR系统的仿真结果表明,该控制器对非线性、时变对象有很好的跟踪控制效果和鲁棒性。

Adaptively controlling algorithm based on neural network of multi-step prediction was proposed for the industrial processes. Developed RBF neural network was used to model the process. The control algorithm trained the neural network controller parameters with performance index based on multi-step predictive error so that adaptive controlled parametens were made to have optimum searching on line. The simulation results of continuous stirred tank reactor （CSTR） expatiate that this controller has good effectiveness and robustness in non-linear featured object and the time-changing one.