摘要
针对IGBT老化失效问题,提出一种基于遗传算法改进的小波神经网络时间序列预测方法。在分析IGBT失效原理的基础上,利用IGBT老化数据集,选取关断瞬时"集电极-发射极"尖峰电压为失效特征参数,采用滑动时间窗法构建训练集与测试集,然后在MATLAB中搭建遗传算法改进的小波神经网络预测模型进行预测,并与传统的小波神经网络预测模型对比分析。试验结果显示,遗传算法改进的小波神经网络预测方均误差为0.017 1,方均根误差为0.130 9,平均绝对误差为0.109 6,分别比传统小波神经网络预测模型降低了0.005 7, 0.020 0, 0.064 0,有效提升了IGBT时间预测的精度。
Aiming at the aging failure of IGBT, an improved wavelet neural network sequentially prediction method based on genetic algorithm was proposed. Based on the analysis of IGBT failure mechanism, with the IGBT aging data, the instantaneous collector emitter peak voltage was selected as the failure characteristic parameter, the training set and test set were constructed by the sliding time window method, and then the wavelet neural network prediction model improved by genetic algorithm was built in MATLAB for prediction, which was compared with the traditional wavelet neural network prediction model. The experimental results show that the mean square error of the improved wavelet neural network is 0.017 1, the root square mean error is 0.130 9, and the average absolute error is 0.109 6, compared with the traditional wavelet neural network prediction model, they are reduced by 0.005 7,0.020 0 and 0.064 0 respectively, which effectively improves the accuracy of IGBT sequentially prediction.
作者
黄柯勋
吴松荣
向碧楠
徐睿
涂振威
HUANG Kexun;WU Songrong;XIANG Binan;XU Rui;TU Zhenwei(Key Laboratory of the Ministry of Education for Maglev Technology and Trains,Chengdu,Sichuan 610031,China;School of Electrical Engineering,Southwest Jiaotong University,Chengdu,Sichuan 611756,China)
出处
《机车电传动》
北大核心
2021年第5期161-166,共6页
Electric Drive for Locomotives
基金
四川省重大科技专项资助项目(20QYCX0095)。
