小波包能量和LM-BP神经网络的发电机机端故障类型识别

Fault Type Identification of Generator Terminal Based on Wavelet Packet Energy and LM-BP Network

为了准确且快速地检测同步发电机出现的各种突然相短路故障,提出了以下检测方法。利用小波包对采集到的相电流进行三层分解和重构。对第三层小波包重构的小波系数进行能量计算并进行归一化处理,然后把处理后的24个值作为BP神经网络的输入数据同时采用LM算法来进行训练和测试。同时与标准BP神经网络的训练效果进行了对比,LM算法优化BP神经网络的训练具有收敛速度快,精确率高的特点。搭建了同步发电机的MATLAB仿真模型对发生各种出现的突然相故障类型进行了仿真,LM-BP神经网络训练和测试的结果表明此方法能够准确地判断发电机所发生的相故障以及故障类型。

In order to accurately and rapidly detect all faults of synchronous generator sudden phase shortcircuit,the following test method is proposed.Make use of wavelet packet to do three layers of reconstructionon collected phase current.Do energy calculation and normalized processing on the wavelet coefficients ofthe three layers of decomposition reconstruction,then put processed twenty-four values as input data to theBP neural network using LM algorithm for training and testing.At the same time compared with the standardBP neural network training effect,LM algorithm to optimize the BP neural network's training effect has the characteristicsof fast convergence rate and high precision rate.Setting up a MATLAB simulation model for thegenerator of all kinds of sudden phase fault is simulated.The training and testing results of LM-BP network showthat this method can accurately judge phase fault and fault type of generator.