基于SSA-LSTM模型的IGBT时间序列预测研究

Research on IGBT Sequentially Prediction Based on SSA-LSTM Model

针对绝缘栅双极型晶体管(IGBT)长工作周期导致的老化失效问题,提出一种基于麻雀搜索算法(SSA)优化长短期记忆(LSTM)网络的IGBT时间序列预测方法。首先分析IGBT疲劳失效的原因,选取某IGBT老化数据集中的集射极峰值电压为失效特征量,进行二次指数滤波以增大数据下降趋势。然后利用Matlab搭建LSTM模型,并采用SSA对网络模型中学习率、隐藏层节点数和训练次数进行寻优以得到最优网络。最后选取常用回归预测性能评估指标对LSTM模型与SSA-LSTM模型预测结果进行对比分析。结果表明,SSA-LSTM模型的预测结果平均绝对误差、均方根误差及平均绝对百分比误差分别降低了0.016%、0.022%和0.202%,证明所提方法预测精度高,可在一定程度上评估IGBT的寿命。

Aiming at the aging failure problem caused by the long operating cycle of insulated gate bipolar transistors(IGBTs),an IGBT sequentially prediction method was proposed which optimizing the long short-term memory(LSTM)network based on the sparrow search algorithm(SSA).Firstly,the causes of IGBT fatigue failure were analyzed,and secondary exponential filtering was performed to increase the downward trend of the data based on the peak collector-emitter voltages in an IGBT aging data set as the failure characteristic quantity.Secondly,the LSTM model was built by Matlab,and the SSA was used to majorize the key parameters such as learning rate,number of hidden layer nodes and number of trainings in the network model to obtain the optimal network.Finally,the commonly used regression prediction performance evaluation indicators were selected to compare and analyze the prediction results of the LSTM and the SSA-LSTM model.The results show that the prediction mean absolute error,root mean square error and mean absolute percentage error of SSA-LSTM are reduced by 0.016%,0.022%and 0.202%respectively.It is proved that the proposed method has the advantage of high prediction accuracy and can evaluate the lifetime of IGBT to a certain extent.