The accurate identification of smart meter(SM)fault types is crucial for enhancing the efficiency of operationand maintenance(O&M)and the reliability of power collectionsystems.However,the intelligent classificati...The accurate identification of smart meter(SM)fault types is crucial for enhancing the efficiency of operationand maintenance(O&M)and the reliability of power collectionsystems.However,the intelligent classification of SM fault typesfaces significant challenges owing to the complexity of featuresand the imbalance between fault categories.To address these issues,this study presents a fault diagnosis method for SM incorporatingthree distinct modules.The first module employs acombination of standardization,data imputation,and featureextraction to enhance the data quality,thereby facilitating improvedtraining and learning by the classifiers.To enhance theclassification performance,the data imputation method considersfeature correlation measurement and sequential imputation,and the feature extractor utilizes the discriminative enhancedsparse autoencoder.To tackle the interclass imbalance of datawith discrete and continuous features,the second module introducesan assisted classifier generative adversarial network,which includes a discrete feature generation module.Finally,anovel Stacking ensemble classifier for SM fault diagnosis is developed.In contrast to previous studies,we construct a two-layerheuristic optimization framework to address the synchronousdynamic optimization problem of the combinations and hyperparametersof the Stacking ensemble classifier,enabling betterhandling of complex classification tasks using SM data.The proposedfault diagnosis method for SM via two-layer stacking ensembleoptimization and data augmentation is trained and validatedusing SM fault data collected from 2010 to 2018 in Zhejiang Province,China.Experimental results demonstrate the effectivenessof the proposed method in improving the accuracyof SM fault diagnosis,particularly for minority classes.展开更多
基金supported by the National Key R&D Program of China(No.2022YFB2403800)the National Natural Science Foundation of China(No.52277118)+1 种基金the Natural Science Foundation of Tianjin(No.22JCZDJC00660)the Open Fund in the State Key Laboratory of Alternate Electrical Power System With Renewable Energy Sources(No.LAPS23018).
文摘The accurate identification of smart meter(SM)fault types is crucial for enhancing the efficiency of operationand maintenance(O&M)and the reliability of power collectionsystems.However,the intelligent classification of SM fault typesfaces significant challenges owing to the complexity of featuresand the imbalance between fault categories.To address these issues,this study presents a fault diagnosis method for SM incorporatingthree distinct modules.The first module employs acombination of standardization,data imputation,and featureextraction to enhance the data quality,thereby facilitating improvedtraining and learning by the classifiers.To enhance theclassification performance,the data imputation method considersfeature correlation measurement and sequential imputation,and the feature extractor utilizes the discriminative enhancedsparse autoencoder.To tackle the interclass imbalance of datawith discrete and continuous features,the second module introducesan assisted classifier generative adversarial network,which includes a discrete feature generation module.Finally,anovel Stacking ensemble classifier for SM fault diagnosis is developed.In contrast to previous studies,we construct a two-layerheuristic optimization framework to address the synchronousdynamic optimization problem of the combinations and hyperparametersof the Stacking ensemble classifier,enabling betterhandling of complex classification tasks using SM data.The proposedfault diagnosis method for SM via two-layer stacking ensembleoptimization and data augmentation is trained and validatedusing SM fault data collected from 2010 to 2018 in Zhejiang Province,China.Experimental results demonstrate the effectivenessof the proposed method in improving the accuracyof SM fault diagnosis,particularly for minority classes.
