{"title":"Research on intelligent evaluation model of transient stability based on K-means grouping","authors":"Xiancheng Ren, Yin Zhang, Feng Wu, H. Yuan, Jinlong Zhang, Haijun Chang","doi":"10.1109/CIEEC54735.2022.9845894","DOIUrl":null,"url":null,"abstract":"In a large number of actual power system operation scenarios, the transient instability scenarios are far smaller than the stable operation scenarios. In the case of imbalanced samples, the rapid transient stability assessment based on machine learning will lead to the learning algorithm preferring to the majority of classes with more samples, thus causing the problem of missing judgment for the transient instability scenarios. This paper proposes an under-sampling method for grouping majority class samples based on K-means. Under the constraint that the proportion of stable samples and unstable samples meets certain conditions, the majority of class stable samples are grouped with the center sample and the sample nearest to the center sample as initial values, and the K-means algorithm is adopted. By regrouping the samples of the majority class successively, multiple training sample subsets composed of unstable samples and multiple groups of stable samples are finally formed. Model training is performed for each training sample set, and the evaluation model of the new mode is judged by the comprehensive distance from the unstable center sample and the stable center sample of each sample subset, and then the transient stability judgment is carried out, so as to improve the accuracy of rapid evaluation under the unbalanced condition of training samples. The effectiveness of the proposed method is verified by an actual power grid example.","PeriodicalId":416229,"journal":{"name":"2022 IEEE 5th International Electrical and Energy Conference (CIEEC)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE 5th International Electrical and Energy Conference (CIEEC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/CIEEC54735.2022.9845894","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
In a large number of actual power system operation scenarios, the transient instability scenarios are far smaller than the stable operation scenarios. In the case of imbalanced samples, the rapid transient stability assessment based on machine learning will lead to the learning algorithm preferring to the majority of classes with more samples, thus causing the problem of missing judgment for the transient instability scenarios. This paper proposes an under-sampling method for grouping majority class samples based on K-means. Under the constraint that the proportion of stable samples and unstable samples meets certain conditions, the majority of class stable samples are grouped with the center sample and the sample nearest to the center sample as initial values, and the K-means algorithm is adopted. By regrouping the samples of the majority class successively, multiple training sample subsets composed of unstable samples and multiple groups of stable samples are finally formed. Model training is performed for each training sample set, and the evaluation model of the new mode is judged by the comprehensive distance from the unstable center sample and the stable center sample of each sample subset, and then the transient stability judgment is carried out, so as to improve the accuracy of rapid evaluation under the unbalanced condition of training samples. The effectiveness of the proposed method is verified by an actual power grid example.