中文摘要：

探索了BP（Back Propagation）神经网络在IGBT模块结温预测中的应用。首先从物理机理角度分析了IGBT模块中饱和压降、集电极电流与温度的内在关系。通过IGBT结温测试试验，获取了三者的关系图，并建立了基于饱和压降的温敏参数法结温预测模型。最后建立了BP神经网络结温预测模型，并与温敏参数法进行了对比分析，证明该方法计算速度更快．误差降低。

英文摘要：

The application of BP neural network method in the prediction of junction temperature of IGBT module is researched. The intrinsic relationship between the saturation voltage drop, collector current and junction temperature is analyzed from the physical mechanism firstly. The curve graph of them is obtained through the test of IGBT temperature, and the junction temperature prediction model of temperature sensitive parameter method based on saturation voltage drop is established. Finally, the temperature prediction model of BP（ Back Propagation） neural network is established. And it is compared and analyzed with the temperature sensitive parameter method, the rapidity and the accuracy of this method are proved.