基于PSO-BP神经网络的IGBT结温在线预测

On-line Prediction of IGBT Junction Temperature Based on PSO-BP Neural Network

针对目前在线监测绝缘栅双极型晶体管(IGBT)结温存在的测量误差大以及检测电路复杂的问题,提出了一种基于粒子群优化(PSO)-反向传播(BP)神经网络的IGBT结温在线提取技术。首先搭建饱和压降平台提取IGBT的饱和压降和集电极电流作为温敏电参数,表征其与IGBT模块结温的关系;然后结合试验数据建立基于PSO-BP算法的结温预测模型,经过训练得到最优网络模型;再将最优网络移植到可编程芯片现场可编程门阵列(FPGA)中,仿真验证FPGA输出结温的准确性;最后搭建结温预测实验平台在线预测IGBT的结温。仿真和实验结果表明,PSO-BP算法的拟合优度达到0.9937,FPGA输出的预测结温能够很好地跟踪实际值,表明所提出方案的准确性和有效性。

To address the problems of large measurement errors and complex detection circuits in on-line monitoring of insulated gate bipolar transistor(IGBT)junction temperature,an on-line extraction technique of IGBT junction tem-perature based on particle swarm optimization(PSO)-back propagation(BP)neural network is proposed.Firstly,the saturation voltage drop platform is built to extract the saturation voltage drop and collector current of IGBT as temper-ature-sensitive electrical parameters to characterize their relationship with the junction temperature of IGBT module.Then the experimental data are combined to establish a junction temperature prediction model based on the PSO-BP algorithm,and the optimal network model is obtained after training.The optimal network is transplanted to the pro-grammable chip field-programmable gate array(FPGA),and the accuracy and realtime performance of FPGA output junction temperature are verified by simulation.Finally,a junction temperature prediction experimental platform is built to realize the junction temperature of the IGBT on-line prediction.Simulation and experimental results show that PSO-BP algorithm has a good fit of 0.9937,and predicted junction temperature output from FPGA can track the actual value well which indicates the accuracy and effectiveness of the proposed scheme.