基于DA-WGAN-SVM的水电机组小样本故障诊断  

Small Sample Fault Diagnosis of Hydroelectric Units Based on DA-WGAN-SVM

作　　者：伍盛金 高金林 赖兴全 王卫玉 魏加达 田波 郑阳 WU Sheng-jin;GAO Jin-lin;LAI Xing-quan;WANG Wei-yu;WEI Jia-da;TIAN Bo;ZHENG Yang(Chongqing Jiangkou Hydropower Co.,Ltd.,SPIC,Chongqing 408506,China;Wuling Power Corporation Ltd.,Changsha 410004,China;School of Power and Mechanical Engineering,Wuhan University,Wuhan 430072,China)

机构地区：[1]国家电投集团重庆江口水电有限责任公司,重庆408506 [2]五凌电力有限公司,湖南长沙410004 [3]武汉大学动力与机械学院,湖北武汉430072

出　　处：《水电能源科学》2024年第11期155-159,共5页Water Resources and Power

基　　金：湖北省科学基金联合基金项目(2022CFD165)。

摘　　要：针对水电机组标定的故障样本偏少的现象,设计了基于DA-WGAN-SVM的水电机组故障诊断模型。首先针对生成对抗网络容易过拟合的问题,采用可微分数据增强实现生成器和判别器的数据增强,提高了生成对抗网络的稳定性,接着利用优化的生成对抗网络实现水电机组故障样本的扩充,最后利用支持向量机对样本进行分类。实例分析验证了该方法的有效性,可为实际工程应用提供参考。Considering the phenomenon of insufficient fault samples for calibration of hydroelectric units,this paper designs a fault diagnosis model for hydroelectric units based on DA-WGAN-SVM.Firstly,in order to solve the problem that it is easy to overfitting the generated countermeasure network,the differential data enhancement is used to enhance the data of the generator and discriminator,which improves the stability of the generated countermeasure network.Sec-ondly,the optimized generated countermeasure network is used to expand the fault samples of hydropower units.Finally,the support vector machine is used to classify the samples to verify the effectiveness of the method,which can provide ref-erence for practical engineering application.

关 键 词：水电机组 故障诊断 生成对抗网络 可微分数据增强 

分 类 号：TV734[水利工程—水利水电工程]