改进型BP神经网络的非线性多模型自适应控制

Multi-model Adaptive Control of Nonlinear Systems Based on Improved BP Neural Network

针对一类参数未知的非线性离散系统,提出一种基于改进型BP神经网络的多模型控制方法。首先将非线性系统表示为线性部分和非线性部分。当非线性部分对系统影响较小时,则直接采用基于固定模型和自适应模型而设计的鲁棒控制器对系统进行控制;而当非线性部分对系统影响较大时,则采用基于改进的BP神经网络的自适应控制。其次,利用切换准则对控制输入进行平滑切换并给出了稳定性证明。最后,仿真结果表明所提方法能提高系统控制品质、减少控制信号的振荡。

As for a class of nonlinear discrete systems with unknown parametersa multi-model control method based on the improved BP neural network is proposed.Firstlythe nonlinear system is expressed with a linear part and a nonlinear part.When the nonlinear part has small impact on the systemthe linear robust controller designed based on the fixed model and the adaptive model is directly used to control the system.When the nonlinear part has large impact on the systemthe adaptive control under the improved BP neural network is adopted.Secondlythe switching criterion is used to smoothly switch the control input and the stability of the system is proved.Finallysimulation results show that the proposed method can improve the system control quality and reduce the oscillation of the control signal.