基于小波包变换和随机森林算法的光伏系统故障分类

针对光伏系统故障分类问题,提出一种小波包变换和随机森林算法相结合的故障分类方法。采集光伏系统的故障电压数据,利用小波包变换对电压信号进行分解,提取各频带能量作为故障特征,将特征样本送入随机森林算法中进行分类。随机森林算法是结合集成学习理论和随机子空间方法的一种算法,可以对多种故障做出准确分类。使用PSCAD/EMTDC搭建独立光伏发电系统,选取12种故障进行模拟,得到600个故障样本,选取其中360个样本用于训练分类器,240个样本用于测试分类器的分类性能。仿真结果表明:该方法可有效辨别光伏系统的12种故障,分类准确率达到97.92%。与RBF神经网络分类器相比,故障分类准确率提高了4.17%,对进一步实现光伏系统故障诊断研究具有重要意义。

A Fault Classification Method of Photovoltaic Systems Based on Wavelet Packet Transform and Random Forest

Aiming at the problem of photovoltaic system fault classification, a fault classification method that combines wavelet packet transform and random forest algorithm is proposed. The fault voltage data of the photovoltaic system are first collected, then the wavelet packet transform is used to decompose the voltage signal, the energy of each frequency band is extracted as the fault feature, and the feature samples are sent into the random forest algorithm for classification. The random forest algorithm is a algorithm that combines ensemble learning theory and random subspace method, which can accurately classify various faults. The independent photovoltaic power generation system is built by PSCAD/EMTDC, 12 types of faults are selected for simulation, 600 samples of fault feature are obtained, among which 360 samples are used to train the random forest classifier, and 240 samples are used to test the classification performance of the classifier. The simulation results show that this method can effectively identify 12 types of faults in the photovoltaic system, and the classification accuracy rate reaches 97.92%. Compared with the RBF neural network, the fault classification accuracy rate is increased by 4.17%, which have important meaning for the further realization of photovoltaic system fault diagnosis research.