【摘要】 针对逆变器功率管开路故障的诊断精度较低问题,对三相电压源型逆变器采用卷积神经网络的方法由现有故障数据训练得到故障识别模型,将不同类别故障对应的逆变器输出侧三相电流信号作为数据集,应用二维卷积神经网络并采用3个通道分别训练三相电流信号,采用Adam优化算法并引入dropout深度学习技巧及自适应学习速率防止模型过拟合,与SVM、KNN、DNN等方法的结果对比表明,该方法可明显提高逆变器功率管开路故障的诊断精度。更多还原

【Abstract】 In order to improve the diagnostic accuracy of the open-switch fault of the inverter, the three-phase voltage source inverter was taken as the research object. The fault recognition model was obtained by using the convolution neural network method to learn the fault data. The three-phase output current signals of the inverter were collected corresponding to different types of faults as a data set. The two-dimensional convolution neural network was applied and three channels were used to train the three-phase current signal. Using the Adam optimization algorithm, the depth learning technique and adaptive learning rate were introduced to prevent over-fitting of the model. The comparison between the proposed method and SVM, KNN, DNN showed that this method obviously improves the diagnostic accuracy of the open-switch fault of inverter.更多还原