S. Sun, Yatong Zhou, Haonuo He, Jingfei He, Yue Chi
{"title":"基于MixMatch的电力负荷数据异常检测","authors":"S. Sun, Yatong Zhou, Haonuo He, Jingfei He, Yue Chi","doi":"10.1109/ICCEAI52939.2021.00010","DOIUrl":null,"url":null,"abstract":"With the development of power industry, electricity has become one of the most important energy sources in our country, related to the lifeline of the country's economy. The electricity system is becoming more and more mature, but abnormal electricity consumption behaviors are also emerging endlessly, causing potential safety hazards in the electricity industry and even the electricity supply system. Considering the lack of abnormal annotations in the electricity load data, this paper proposes a semi-supervised electricity load data anomaly detection method based on MixMatch. Firstly, data cleaning of electricity load data is used to remove incorrect data. Secondly, Convolutional Autoencoder (CAE) is used to extract its time-domain and frequency-domain features separately, and the features are combined through feature fusion. Thirdly, the Borderline Synthetic Minority Oversampling Technique (Borderline-SMOTE) is used to solve the problem of data imbalance. The MixMatch semi-supervised algorithm is used to label the abnormal data to realize the anomaly detection of the electricity load data. Finally, this paper uses the k-means clustering and T-Stochastic neighbour Embedding (T -SNE) to classify the abnormal data and visualize the data. The experimental results show that, compared with traditional machine learning methods, the proposed method has a significant improvement on AUC.","PeriodicalId":331409,"journal":{"name":"2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Anomaly detection of electricity load data based on MixMatch\",\"authors\":\"S. Sun, Yatong Zhou, Haonuo He, Jingfei He, Yue Chi\",\"doi\":\"10.1109/ICCEAI52939.2021.00010\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"With the development of power industry, electricity has become one of the most important energy sources in our country, related to the lifeline of the country's economy. The electricity system is becoming more and more mature, but abnormal electricity consumption behaviors are also emerging endlessly, causing potential safety hazards in the electricity industry and even the electricity supply system. Considering the lack of abnormal annotations in the electricity load data, this paper proposes a semi-supervised electricity load data anomaly detection method based on MixMatch. Firstly, data cleaning of electricity load data is used to remove incorrect data. Secondly, Convolutional Autoencoder (CAE) is used to extract its time-domain and frequency-domain features separately, and the features are combined through feature fusion. Thirdly, the Borderline Synthetic Minority Oversampling Technique (Borderline-SMOTE) is used to solve the problem of data imbalance. The MixMatch semi-supervised algorithm is used to label the abnormal data to realize the anomaly detection of the electricity load data. Finally, this paper uses the k-means clustering and T-Stochastic neighbour Embedding (T -SNE) to classify the abnormal data and visualize the data. The experimental results show that, compared with traditional machine learning methods, the proposed method has a significant improvement on AUC.\",\"PeriodicalId\":331409,\"journal\":{\"name\":\"2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-08-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICCEAI52939.2021.00010\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICCEAI52939.2021.00010","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Anomaly detection of electricity load data based on MixMatch
With the development of power industry, electricity has become one of the most important energy sources in our country, related to the lifeline of the country's economy. The electricity system is becoming more and more mature, but abnormal electricity consumption behaviors are also emerging endlessly, causing potential safety hazards in the electricity industry and even the electricity supply system. Considering the lack of abnormal annotations in the electricity load data, this paper proposes a semi-supervised electricity load data anomaly detection method based on MixMatch. Firstly, data cleaning of electricity load data is used to remove incorrect data. Secondly, Convolutional Autoencoder (CAE) is used to extract its time-domain and frequency-domain features separately, and the features are combined through feature fusion. Thirdly, the Borderline Synthetic Minority Oversampling Technique (Borderline-SMOTE) is used to solve the problem of data imbalance. The MixMatch semi-supervised algorithm is used to label the abnormal data to realize the anomaly detection of the electricity load data. Finally, this paper uses the k-means clustering and T-Stochastic neighbour Embedding (T -SNE) to classify the abnormal data and visualize the data. The experimental results show that, compared with traditional machine learning methods, the proposed method has a significant improvement on AUC.