基于三序分量法和PNN的配电网不对称故障类型识别

Asymmetric Fault Type Recognition of Distribution Network Based on Three Sequence Component Method and PNN

针对配电网故障类型识别率低的问题,提出一种基于正、负、零三序分量和概率神经网络(PNN)的配电网故障类型识别方法。首先,利用Matlab软件对10 kV配电网发生各类不对称故障进行仿真,将产生的故障电流用对称分量法进行分解,然后得到有差异的正、负、零序电流;再用这3种分量作为特征量,代入PNN进行训练;最后,将现有的故障特征量输入训练完成且具有识别功能的PNN网络中,达到故障识别及分类的目的。仿真结果表明:在不同的故障合闸角、过渡电阻以及负荷有重大变化等情况下,三序分量法与负序分量法在区分单相接地、两相短路接地、两相相间短路等不对称故障时,前者区分准确度更高,对进行事故分析和故障选相等具有重要意义。

Aimingat the problem of low recognition rate of distribution network fault types,a method of distribution network fault type recognition based on positive,negative,zero three-sequence components and probabilistic neural network(PNN)is proposed.Firstly,all kinds of asymmetric faults in 10 kV distribution network are simulated by using Matlab software.The fault current is decomposed by symmetric component method,and then the positive,negative and zero sequence currents with different characteristics are obtained.Then,the three components are used as feature quantities,which are brought into PNN for training.Finally,the existing fault features are input into the PNN network with recognition function after training,so as to achieve the purpose of fault identification and classification.The simulation results show thatthe three-sequence component method and the negative-sequence component method have higher accuracy in distinguishing asymmetric faults such as single-phase grounding,two-phase short-circuit grounding and two-phase short-circuit under different fault closing angles,transition resistances and significant load changes,which is of great significance for accident analysis and fault phase selection.