负序电流是感应电机定子故障的主要故障特征量之一，但是在实际电机中，供电电压不平衡，电机制造和装配误差造成的先天不平衡、负载的大小变化等都会影响负序电流的大小，因此不考虑这些因素的影响而单纯从负序电流的大小来诊断定子故障及故障的严重程度容易出现误判。当定子故障发生时，总的负序电流是由负序电压产生的负序电流分量、电机先天不平衡产生的负序电流分量以及定子绕组故障引起的负序电流分量的相量和构成的。该文通过派克矢量的旋转变换，可以在大量谐波和噪声存在的情况下，求得电流和电压的基波正序和负序相量，并通过负序阻抗和径向基函数神经网络排除负序电流中其他非故障因素的影响，得到负序电流中与定子故障相对应的分量，并以此为依据进行故障诊断。实验结果表明，该方法是一种可行、有效的方法。

The negative sequence current is one of the indicators of induction motor stator fault. But in a real case, the unbalanced supply voltage sources and the inherent asymmetries in the motor and variation of loads cause a change of the negative sequence current. So it may go to a wrong result if these non-idealities are not taken into properly account. As a stator fault occurs, the phasor of total negative sequence current is the summation phasors of the negative sequence currents contributed by stator fault, the negative sequence voltage and inherent asymmetries of the motor. This paper applied Park’s vector rotating transformation to calculate the phasors of fundamental positive and negative sequence voltage and current against a background of a lot of harmonics and noise existed in the signal. Then the portion of the negative sequence current caused by the factors other than stator fault is taken out by a negative sequence impedance and a radial basis function (RBF) neural network. The residual negative sequence current is entirely caused by stator fault, and it is used as fault indicator. The proposed technique is verified by applying it to diagnose the stator fault in a three-phase induction motor.

国家自然科学基金项目(50677069)。