摘要
针对IGBT寿命预测问题,提出了基于思维进化算法(MEA)优化的反向传播(MEA-BP)神经网络算法的IGBT结温预测算法模型,确定了集电极-发射极关断电压尖峰峰值为失效预测依据,构建了遗传算法(GA)优化的BP(GA-BP)神经网络算法以及经典BP神经网络算法寿命模型作为对比模型,采用均方误差、平均绝对误差、最大相对误差作为各模型预测性能的评估指标。预测结果表明,基于MEA神经网络的IGBT寿命预测模型均方误差为0.150%,平均绝对百分误差为0.36%,可以更好地实现IGBT寿命的预测。
Aiming at the problem of IGBT life prediction,a model of IGBT junction temperature prediction algorithm based on back propagation(mea-bp)neural network algorithm optimized by thought evolutionary algorithm(MEA)is proposed in this paper.The peak value of Collector Emitter turn off voltage is determined as the basis of failure prediction.The GA-BP neural network algorithm optimized by genetic algorithm(GA)and the life model of classical BP neural network algorithm are constructed as comparison models.Mean square error,mean absolute error and maximum relative error are used as evaluation indexes of prediction performance of each model.The prediction results show that the mean square error and the average absolute percentage error of the IGBT life prediction model based on mind evolutionary algorithm neural network are 0.150%and 0.36%,respectively,which can better realize the IGBT life prediction.
作者
王新春
李锦涛
WANG Xinchun;LI Jintao(Inner Mongolia University of Science and Technology,Baotou 014010,China)
出处
《电工技术》
2021年第18期116-119,123,共5页
Electric Engineering
关键词
IGBT
思维进化算法(MEA)
神经网络
寿命预测
insulated gate bipolar transistor(IGBT)
mind evolutionary algorithm(MEA)
BP neural network
life prediction
