基于BP和RBF神经网络的水轮机综合特性曲线的拟合

Hydroturbine synthetic characteristic curve fitting based on BP and RBF neural network

关于大波动过渡过程中水轮机特性曲线的处理有很多种方法,其中有用神经网络方法来训练相关数据形成网络从而预测待求数据点,其中BP和RBF两种网络为较为常用的神经网络方法。本文通过学习Matlab中神经网络工具箱的调用,比较分析了BP和RBF两种神经网络在水轮机特性曲线拟合上的区别,试图找出BP网络隐含层神经元个数对其拟合精度的影响,以及随着其神经元个数增加而拟合度增加的情况。

There are many fitting methods for the hydroturbine characteristic curve under the transiting process of acute fluctuation. The neural network method is used to forecast the solving data points by training related data to form the network, the BP and RBF network are more common neural network methods. Through the call study of neural network toolbox in the Matlab software, the paper compares and analysis the difference of hydroturbine characteristic curve fitting between the BP and RBF neural network, tries to find the influence of neuron number in hidden layer of BP network on the curve fitting precision and analyzes the relationship of fitting accuracy degree improvement with the increasing of the neuron number.