汽车工程 ›› 2022, Vol. 44 ›› Issue (1): 142-152.doi: 10.19562/j.chinasae.qcgc.2022.01.017

所属专题: 新能源汽车技术-电驱动&能量管理2022年

• • 上一篇    

基于卷积神经网络的逆变器故障诊断方法

于海,邓钧君(),王震坡,孙逢春   

  1. 1.北京理工大学,电动车辆国家工程实验室,北京 100081
    2.北京电动车辆协同创新中心,北京 100081
  • 收稿日期:2021-08-16 修回日期:2021-11-04 出版日期:2022-01-25 发布日期:2022-01-21
  • 通讯作者: 邓钧君 E-mail:dengjunjun@bit.edu.cn
  • 基金资助:
    国家重点研发计划(2019YFB1600800)

Inverter Fault Diagnosis Method Based on Convolutional Neural Network

Hai Yu,Junjun Deng(),Zhenpo Wang,Fengchun Sun   

  1. 1.Beijing Institute of Technology,National Engineering Laboratory for Electric Vehicles,Beijing 100081
    2.Collaborative Innovation Center for Electric Vehicles in Beijing,Beijing 100081
  • Received:2021-08-16 Revised:2021-11-04 Online:2022-01-25 Published:2022-01-21
  • Contact: Junjun Deng E-mail:dengjunjun@bit.edu.cn

摘要:

针对车用永磁同步电机驱动系统在长期运行过程中存在的逆变器故障风险,本文中提出了一种基于卷积神经网络的逆变器故障诊断方法。首先,对三相定子电流数据进行标幺化和单位电周期电流数据筛选处理,降低电机驱动系统变转矩、变转速工况对故障诊断效果的影响;然后,结合卷积神经网络,发挥其提取故障特征和抗噪优势,实现逆变器故障诊断。在MATLAB/Simulink环境中搭建了车用永磁同步电机驱动系统模型,通过故障注入仿真数据和实验数据构建用于卷积神经网络训练的数据集,验证了所提出故障诊断方法的有效性。同时,探究了训练样本数据集稀疏化和电机型号差异化情况的适用性。仿真结果表明:在噪声数据和稀疏数据条件下,本文中提出的故障诊断方法具有很好的鲁棒性和普适性。

关键词: 逆变器, 三相定子电流, 卷积神经网络, 故障诊断

Abstract:

For the risk of inverter failure in the permanent magnet synchronous motor (PMSM) drive system of the electric vehicles during long-term operation, a fault diagnosis method based on convolutional neural network (CNN) is proposed in this paper. Firstly, three-phase stator current data is normalized and filtered to an electric cycle current data, reducing the impact of variable torque and variable speed conditions of the motor drive system on the fault diagnosis effect. Then, the CNN, for the advantage of fault feature extraction and noise immunity, is used for inverter fault diagnosis. In MATLAB/Simulink, the PMSM drive system model of the electric vehicle is built. The fault injection simulation and experimental data is used to construct the dataset for the CNN, and the effectiveness of the proposed fault diagnosis method is verified. In addition, the applicability of the method is explored in the case of training sample dataset sparsification and motor model differentiation.The simulation results show that the fault diagnosis method proposed in this paper is robust and universal under the conditions of noise data and sparse data.

Key words: inverter, three-phase stator current, convolutional neural network, fault diagnosis