基于LSTM与电气参数的电机状态监测方法  

Condition Monitoring Method of Motor Based on LSTM and Electrical Parameters

作　　者：杨磊[1] 雷成[1] 李亮[1] 王镜淇 Yang Lei;Lei Cheng;Li Liang;Wang Jingqi(CNNC Jiangsu Nuclear Power Corporation,Lianyungang,Jiangsu 222000,China)

机构地区：[1]江苏核电有限公司,江苏连云港222000

出　　处：《机电工程技术》2023年第7期164-169,共6页Mechanical & Electrical Engineering Technology

基　　金：电站重点科研项目(JNPC-KY-201933)。

摘　　要：针对现有的电机状态监测方法因状态曲线反复穿越报警线,导致状态预警分析准确率低的问题,提出一种长短期记忆网络(LSTM)与工艺参数相结合的状态监测方法。该方法首先采用主成分分析(PCA)对电机正常运行状态的电流、电压信号特征矩阵进行特征降维,再借助LSTM完成状态预测模型构建,基于历史运行数据实现对当前阶段运行待测数据状态值的预测,最终通过对比实测值与预测值的差异度完成电机运行状态的判断。搭建电机故障实验台对提出方法进行验证,实测数据分析结果表明,所提出的方法可有效实现针对电机的状态监测,与振动参数分析的对比结果显示,该方法具备较好的模型稳定性及准确性,有效解决了状态监测模型误报、漏报的问题,实现了对电机运行状态的准确监测。Aiming at the problem of low accuracy of condition warning analysis due to the state curve crossing alarm line repeatedly,a condition monitoring method combining long short-term memory network(LSTM)and process parameters is proposed.In the method,principal component analysis(PCA)is used to reduce the characteristic dimension of the current and voltage signal characteristic matrix of the normal running state of the motor,and then the state prediction model is constructed with the help of LSTM,and the state value of the data to be measured in the current running stage is predicted based on the historical running data.Finally,the running state of the motor is judged by comparing the difference between the measured value and the predicted value.A motor fault test platform is built to verify the proposed method.The analysis results of the measured data show that the proposed method can effectively realize the condition monitoring of the motor.The comparison results with the analysis of vibration parameters show that the proposed method has better model stability and accuracy,effectively solves the problem of false positives and missing positives of the condition monitoring model,and realizes the accurate monitoring of the running state of the motor.

关 键 词：电机 状态监测 电流 电压 LSTM 

分 类 号：TH165.3[机械工程—机械制造及自动化] TM343.2[电气工程—电机]