【摘要】 研究了一种基于神经网络预测模型的动态矩阵控制(DMC)算法,首先利用多层前馈神经网络辨识对象模型,同时预测对象的未来输出,然后用DMC算法进行滚动优化和反馈校正。通过对一类化工反应器非线性系统的仿真,结果表明该算法在非线性对象的任意工作点都可以通过神经网络辨识获得工作点附近的近似线性模型,具有较好的实时性。更多还原

【Abstract】 A DMC algorithm based on neural networks predictive model is researched.First multilayer feed-ahead neural networks are employed to identify model of the plant and produce predictive output.Then DMC algorithm is used in receding horizon and revising feedback.After the proposed algorithm is applied to simulate a class of nonlinear chemical reactor control systems,the result demonstrates that the approximate linear model near work spot can be achieved in any work spots of nonlinear plant by neural networks identification and it has better real time.更多还原