基于PSO-BP神经网络的SiC MOSFET模块寿命预测方法研究与实现  

Research and Implementation of Life Prediction Method for SiC MOSFET Module Based on PSO-BP Neural Network

作　　者：毛明波 孟昭亮[1,2,3] 高勇 杨媛[2] MAO Mingbo;MENG Zhaoliang;GAO Yong;YANG Yuan(School of Electronics and Information,Xi’an Polytechnic University,Xi’an 710699,China;School of International Engineering,Xi’an University of Technology,Xi’an 710048,China;Power Electronics Division of CRRC Yongji Electric Co.,Ltd.,Xi’an 710000,China)

机构地区：[1]西安工程大学电子信息学院,西安710699 [2]西安理工大学国际工学院,西安710048 [3]中车永济电机有限公司电力电子事业部,西安710000

出　　处：《电源学报》2025年第1期229-235,258,共8页Journal of Power Supply

摘　　要：针对目前碳化硅金属氧化物半导体场效应晶体管Si CMOSFET(siliconcarbidemetal-oxide-semiconductor field-effect transistor)实际工况中在线寿命预测难度大的问题,提出1种基于粒子群优化-反向传播PSO-BP(particle swarm optimization-back propagation)神经网络的SiC MOSFET模块寿命预测数字化实现方法。首先,利用导通压降平台提取Si CMOSFET的导通压降作为温敏电参数,建立基于实验数据的结温预测方案;其次,利用功率循环加速老化实验平台,提取老化特征数据,建立基于PSO-BP神经网络的寿命预测方案;然后,将结温预测方案与寿命预测方案移植到可编程阵列逻辑中,实现SiC MOSFET寿命预测数字化;最后,设计了验证电路。实验表明,数字化显示的结温与真实结温的误差为4.73℃,与真实寿命次数的误差百分比为4.1%,证明所提寿命预测方法得到了数字化实现,并能够准确预测SiC MOSFET模块的寿命次数。To solve the difficulty in online life prediction of silicon carbide metal-oxide-semiconductor field-effect transistor(SiC MOSFET)under practical working conditions,a digital implementation method for SiC MOSFET module life prediction based on particle swarm optimization-back propagation(PSO-BP)neural network was proposed.First,the saturation voltage drop of SiC MOSFET was extracted by a saturation voltage drop platform as the temperature-sensitive electric parameter,and a junction temperature prediction scheme based on experimental data was established.Second,a life prediction scheme based on PSO-BP neural network was established by using a power cycling accelerated aging experimental platform to extract the aging characteristic data.Third,the junction temperature prediction scheme and life prediction scheme were transplanted to field programmable gate array to realize the digitization of SiC MOSFET life prediction.Finally,a circuit was designed to verify the proposed method.Experimental results show that the error between the digital junction temperature and real junction temperature was 4.73℃,and the percentage of error between the predicted life times and real life times was 4.1%,which proves that the proposed life prediction method is realized digitally and can accurately predict the life times of SiC MOSFET module.

关 键 词：SiC MOSFET 粒子群优化-反向传播 寿命预测 数字化 

分 类 号：TN386[电子电信—物理电子学]