随机森林模型在IGBT器件状态监测中的应用  

Application of Random Forest Model in IGBT Module Condition Monitoring

作　　者：周荔丹 李璟 姚钢[2] 王杰[2] ZHOU Li-dan;LI Jing;YAO Gang;WANG Jie(Shanghai University of Electrical Power,Shanghai 200090,China;不详)

机构地区：[1]上海电力大学,电气工程学院,上海200090 [2]上海交通大学,电子信息与电气工程学院,上海200240

出　　处：《电力电子技术》2023年第8期133-136,共4页Power Electronics

基　　金：国家自然科学基金(52077135)。

摘　　要：针对绝缘栅双极型晶体管(IGBT)模块老化故障频发的问题,这里将优化随机森林模型应用于IGBT老化故障诊断系统中以实现IGBT器件的状态监测和老化故障预测。首先选定老化故障特征参数并进行数据预处理,建立老化故障诊断数据集。其次,建立传统随机森林模型,并在此基础上通过基评估器参数寻优、模型框架参数寻优及袋装法的处理方法对传统随机森林模型进行优化,形成优化随机森林模型。在交叉验证的基础上完成模型训练,最后通过多种评估指标对所提模型和其他模型在IGBT老化故障诊断数据集上的预测效果进行评估,最终优化后的模型拟合情况较好,训练集曲线和测试集曲线之间的误差为1.19%,在测试集上的预测精度可达98.81%,验证了优化随机森林模型应用于IGBT状态监测系统中的可行性和精确性。In order to realize the condition monitoring and aging fault prediction of insulated gate bipolar transistor(IGBT),an optimized random forest model is applied to IGBT aging fault diagnosis system.Firstly,the characteristic parameters of aging fault are selected and the data is preprocessed to establish the aging fault diagnosis data set.Secondly,the traditional random forest model is established,and on this basis,the traditional random forest model is optimized through the parameter optimization of the base evaluator,the parameter optimization of the model frame and the bagging method.The optimized random forest model is formed.Model training is completed on the basis of cross-validation.Finally,the prediction effect of this model and other models on IGBT aging fault diagnosis data set are evaluated by various evaluation indexes.The optimized model fits well,the error between the training set curve and the test set curve is 1.19%,and the prediction accuracy on the test set could reach 98.81%.The feasibility and accuracy of the optimized stochastic forest model applied to IGBT condition monitoring system are verified.

关 键 词：随机森林模型 绝缘栅双极型晶体管 状态监测 老化故障 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]