基于MEEMD与概率神经网络的贯通式同相牵引直接供电系统故障识别  

Fault Identification of Traction Network Based on MEEMD and Probabilistic Neural Network

作　　者：蔡潇 陈仕龙[1] 毕贵红[1] 王泽超 庄启康 CAI Xiao;CHEN Shilong;BI Guihong;WANG Zechao;ZHUANG Qikang(School of Electric Power Engineering,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学电力工程学院,云南昆明650500

出　　处：《电力科学与工程》2020年第1期62-69,共8页Electric Power Science and Engineering

基　　金：国家自然科学基金资助项目(51767012)。

摘　　要：针对贯通式同相牵引直接供电系统可能发生的雷击故障、雷击干扰和接地故障3种扰动进行建模分析和识别研究。在牵引网仿真模型的基础上,通过实验得到3种扰动的暂态特征。根据以上故障提出了改进总体平均经验模态分解(Modified Ensemble Empirical Mode Decomposition,MEEMD)与概率神经网络(Probabilistic Neural Network,PNN)结合的智能识别方法。MEEMD分解故障暂态电流信号得到本征模态函数(Intrinsic Mode Function,IMF),分别用样本熵和排列熵提取IMFs分量特征,结合PNN进行故障识别,通过实验看出,基于MEEMD排列熵与PNN结合的智能识别方法能较好地识别牵引网的3种故障。The modeling analysis and identification of three kinds of disturbance,that is,lightning strike failure,lightning strike interference and ground fault,which may occur in the co-phase traction direct power supply system are carried out.Based on the traction network simulation model,the transient characteristics of three disturbances are obtained through experiments.According to above faults,an intelligent identification method combining improved population modified ensemble empirical mode decomposition(MEEMD)with probabilistic neural network(PNN)is proposed.MEEMD decomposes the transient current signal to obtain the intrinsic mode function(IMF),extracts the IMFs component features with sample entropy and permutation entropy,and combines PNN for fault identification.Simulation results illustrated that the intelligent recognition method based on MEEMD permutation entropy and PNN can identify three kinds of faults of traction network better.

关 键 词：贯通式同相牵引供电系统 MEEMD 概率神经网络 IMF 故障识别 

分 类 号：TM73[电气工程—电力系统及自动化]