采用负载端电压预测串联故障电弧

Series arc fault prediction by using load side voltage

为研究串联故障电弧的检测方法,针对三相电动机及变频器负载开展了串联故障电弧实验.首先采用小波包对负载端电压进行4层分解,并建立5~16节点系数的极限学习机(Extreme learning machine, ELM)预测模型;然后利用建立的ELM预测模型对10个周波的小波包节点系数进行预测,分别计算每个周波预测残差的平均值并排序,取4个最小值的均值作为故障电弧特征;最后选取正常状态下特征的最大值和故障状态下特征的最小值的均值为故障电弧检测的公共阈值.结果表明:上述方法可有效检测三相电动机及变频器负载回路中的故障电弧,同时可排除谐波及暂态干扰.

To study detection method of series arc fault,series arc fault experiments in three-phase motor and inverter load were carried out.Firstly,the load side voltage was decomposed into four layers by using wavelet packet.And an extreme learning machine(ELM)prediction model with 5~16 node coefficients was established.Then,the wavelet packet node coefficients of ten cycles were predicted with the established ELM prediction model.The average value of prediction residual error of each cycle was calculated and sorted.And the mean value of four minimum values was used as fault arc feature.Finally,the average of the maximum feature value in normal state and the minimum feature value in arc fault state was used as common threshold of arc fault detection.The results show that the proposed method can effectively detect serial arc fault in the three-phase motor and inverter load loop even with harmonic or transient interference.