基于神经网络的非线性预测函数控制

Nonlinear predictive functional control based on a neural network

给出了一种新的神经网络多步预估器结构 ,建立了 CSTR过程的人工神经元网络的动态模型 ,并在此基础上提出了基于人工神经元网络模型的非线性预测函数控制算法 .给出了非线性预测函数控制的具体实施步骤 .计算机仿真表明 ,人工神经元网络模型的精度已满足预测控制的需要 ,该控制系统比常规 PID控制器具有更好的控制效果 .

An Artificial Neural Network (ANN) is an adequate tool for modeling nonlinear systems and can be applied straightforward in the predictive functional control which belongs to the classical family of model predictive control. New structure of ANN multi-step prediction that is different from cascade or parallel is given, at the same time, the nonlinear predictive functional control using the ANN model is developed for control of high-nonlinear system. The performance of this strategy is evaluated by applying it to a Continuous Stirred Tank Reactor(CSTR). The results illustrate that the NPFC using ANN model is more effective for control nonlinear system than PID control.