基于GRU和ARIMA混合模型的IGBT失效预测

IGBT Failure Prediction Based on GRU and ARIMA

针对现有绝缘栅双极型晶体管(Insulated Gate Bipolar Transistor,IGBT)失效特征预测误差较大的问题,构建一种基于门控循环单元(Gated Recurrent Units,GRU)和自回归整合移动平均模型(Autoregressive Integrated Moving Average model,ARIMA)的门控自回归混合失效预测模型。混合预测模型利用全连接层连接门控循环单元和自回归整合移动平均模型构建成。实验结果表明,门控自回归混合失效预测模型在均方误差、平均绝对误差和最大相对误差上的误差对比现存方法平均减少约12.94%。

A gated autoregressive hybrid failure prediction model based on Gated Recurrent Units(GRU)and Autoregressive Integrated Moving Average model(ARIMA)was constructed to address the issue of significant prediction errors in the failure characteristics of existing Insulated Gate Bipolar Transistors(IGBT).The hybrid prediction model is constructed using a fully connected layer connected gated loop unit and an autoregressive integrated moving average model.The experimental results show that the error of the gated autoregressive hybrid failure prediction model in mean square error,average absolute error,and maximum relative error is reduced by an average of about 12.94%compared to existing methods.