用于非线性加热炉的神经网络预测控制器

摘要

将预测控制的优化思想与神经网络精确描述非线性和不确定性动态过程的特性有机结合，提出了可用于非线性加热炉的直接优化的神经网络预测控制方法，该方法是采用离线训练的神经网络，通过在线反馈校正，分段优化控制量，可使最优预测输出逼近参考轨迹；基于李雅普诺夫方法，讨论了闭环系统的稳定性，得到闭环系统局部渐进稳定的一个充分条件；讨论了加权系数h与λ对系统的影响；仿真结果表明了该方法的有效性、准确性和鲁棒性。%In this paper, the optimization methods of neural networkpredictive control is put forward, in witch neural network is used to be as a predictive model, and LM methods and data measured practically are adopted to train the neural network off-line. As a result,optimized predictive output can approach the reference track through on-line feedback correction and piecewise optimizing control. Further more, a Lyapunov based closed-loop stability analysis is discussed and a sufficient condition for local asymptotic stability is derived.Weight h and λ are discussed .The results of simulation show that this method is simple, efficient, precise and robust.