基于IPSO-BPNN算法的开关磁阻发电机实验建模

Experimental Modeling of Switched Reluctance Generators Based on IPSO-BPNN

文章实现了一种利用改进的粒子群算法优化BP神经网络(IPSO-BPNN)的建模方法,建立了SRG的非线性模型。该方法利用了BP神经网络较强的非线性处理能力和逼近能力,改进粒子群算法的引入克服了BP神经网络容易陷入局部最优及初值敏感的缺点。建模的实验数据采用间接测量法采集,分为训练样本与测试样本两个集合。建模效果表明IPSO-BP神经网络的泛化能力很强,可以近乎完美地表达SRG的磁链和转矩特性。

This paper implemented a modeling method with the experimental data by improved particle swarm optimization and BP neural network algorithm, and a non-linear model of switched reluctance generator（SRG） was built. The method made full use of the strong nonlinear approximation ability of BP neural network. The improved particle swarm optimization algorithm was combined with BP neural network, which overcame the BP neural network＇s shortcoming of being vulnerable to local optimum and of initial value being sensitive. The modeling experimental data was herborized using an indirect measurement, and it was divided into two collections ： training and test samples. The efforts suggest that the IPSO-BP neural network model has a strong generalization ability,and it can perfectly express the flux and torque characteristics of SRG.