基于神经网络的模型参考自适应控制方法

A Model Reference Adaptive Control Method Based on Neural Network

针对传统基于模糊神经网络的模型参考自适应控制方法的一些不足,提出了一种基于 BP 神经网络的模型参考自适应控制结构,并对所使用的 BP 网络学习算法进行了分析改进。对比分析采用传统自适应方法和改进的自适应方法时,不同的控制仿真结果表明,改进后的方法可以有效地抑制神经网络的"过学习"现象,减小了对神经网络辨识器精度的依赖程度,改进效果显著。

In regard to limitations of the traditional model reference adaptive control method based on fuzzy neural network, a model reference adaptive control structure based on BP neural network is presented, moreover the BP network learning arithmetic is analyzed and improved. Comparison of the control simulation result adopted in the traditional adaptive method with that in the improved adaptive method indicates that the improved method can effectively depress the “over-learning” phenomenon of neural network and reduce the dependence on the precision of neural network identifying system, and the improved effect is notable.