基于Park变换和DRNN的定子绕组匝间故障诊断方法

Park Vector and DRNN based Detection Method for Stator Winding Turn Fault

采用定子电流信号检测方法诊断三相交流电机定子绕组匝间短路故障时,会受到电网电压不对称和负载变化等因素的影响,为克服这一缺陷,提出了基于派克变换和对角递归神经网络(DRNN)的定子绕组匝间故障诊断方法.该方法根据派克变换得到三相电流派克矢量模的轨迹变化,通过频谱分析提取故障严重度特征因子.为进一步确定短路绕组的匝数,综合考虑负载、三相输入电压不平衡度的变化情况,构建基于DRNN的短路匝数诊断模型.根据此方法,构建了试验系统并进行了匝间短路试验,试验结果证明:基于Park变换和DRNN的诊断方法,不但在稳态工况下可精确确定定子绕组短路故障的严重度及匝数,而且在电机启动、负载、电压不平衡动态变化时,取得比前馈神经网络(FFNN)故障诊断模型更好的诊断结果.

Voltage unbalance and load changes influence the current signals for diagnosing interturn short circuit in stator winding of three phase induction motors. To solve this problem, a park vector and diagonal recurrent neural network （DRNN） based detection method for stator winding turn fault was presented. When inter-turn short circuit occurred in the stator winding of three phase induction motors, the current park vector module would change and serious .faults would appear. In order to detect inter-turns accurately, the fault factor was drawn by spectrum analysis and a model which took load and three phase voltage unbalance into considera- tion was constructed based on DRNN. Experiment results have shown that the Park Vector and DRNN based detection method determines the shorted turns exactly in different operating states and is more effective than the feed-forward neural network（FFNN） based detection model when motor starts up and voltage unbalance and load changes.