基于PSO-BP算法的IGBT结温预测方法分析

Analysis of IGBT Junction Temperature Prediction Method Based on PSO-BP Algorithm

针对IGBT结温难以直接测量的缺陷,阐述IGBT结温测量平台,选取不同IC的VCE温敏参数预测结温,分别离线采集实验数据,并分别搭建基于BP神经网络和PSO-BP算法的结温预测模型。实验结果表明,采用PSO-BP算法预测IGBT结温较BP神经网络更接近实测数据,误差更低。

Aiming at the defect that IGBT junction temperature is difficult to measure directly,this paper describes IGBT junction temperature measurement platform,selects Vce temperature sensitive parameters of different Ics to predict junction temperature,collects experimental data offline respectively,and builds junction temperature prediction models based on BP neural network and PSO-BP algorithm respectively.The experimental results show that the predicted IGBT junction temperature using PSO-BP algorithm is closer to the measured data than BP neural network,and the error is lower.