多层前馈神经网络及其在开关变换器中的应用

Multilayer feedforward neural network and its application for switching converters

多层前馈神经网络具有多层结构、可鉴别神经元特性函数和误差反传算法等三个要素。本文从多层前馈神经网络的结构和原理出发,分析了多层前馈神经网络的非线性函数映射能力,以此为基础,从而可以实现其在系统辨识和非线性控制上的作用。另外,以Boost变换器的神经网络辨识器和控制器的设计为例,探讨了开关变换器的数学建模方法和多层前馈神经网络在其中的应用,使得各种线性和非线性控制方法均可以利用多层前馈神经网络来实现,并具有统一的设计和训练措施。

Multilayer feedforward neural network(MFNN) contains three factors,i.e.,multilayer structure,identifiable neuron characteristic function and back-propagation Algorithm.According to the structure and the mechanism of MFNN,this paper analyzes the nonlinear function mapping capabilities.Then the system can achieve the role of its identification and nonlinear control based on MFNN.In addition,the paper discusses the mathematical modeling and multi-layer neural networks application of the converter by using the ...