一种BP MFN逆模型补偿的复合控制结构方案研究

Study of Compound Control Structure Scheme Based on Offset of BP MFN (Multilayer Feedforward Network) Inverse Model

由于对结构未知和不确定的非线性系统还没有形成一种通用有效的辨识和控制方法,为此首先对非线性系统逆模型辨识和控制的结构方案进行分析,然后基于复合控制思想,对基于神经网络的非线性系统逆模型补偿的复合控制结构方案进行研究。设计了一种基于BP MFN(Multilayer Feedforward Network)逆模型补偿的复合控制结构方案,并基于不同BP MFN逆模型结构进行了仿真。仿真结果显示,基于神经网络的非线性系统逆模型补偿的复合控制结构方案是有效的,且在满足辨识建模精度要求前提下,采用相对简单的BP MFN逆模型结构,对提高逆模型的泛化能力和非线性系统的控制效果是有益的。

Because of the universally and effectively method to identify and control for nonlinear system not being formed, analysis was done to identification and control structure scheme of inverse model based on Artificial Neural Network （ANN） for nonlinear system. Used compound control principle, research of the compound control structure scheme based on offset of ANN＇ s inverse model for nonlinear system was done, and the compound control scheme based on offset of the BP MFN＇ s （ Multilayer Feedforward Network） inverse model was designed, and simulation researches were done with different architecture schemes of the BP MFN＇s inverse model. Simulation results show, the compound control structure scheme designed is effective, and the relatively simple network architecture can be raised generalization ability of the BP MFN＇s inverse model, and get well control effect to nonhnear system.