融合连续域蚁群算法One-Class SVM的电力离群用户检测  

Outlier user detection based on continuous domain ant colony algorithm One-Class SVM

作　　者：郭玮 Guo Wei(Hainan Power Grid Co.,Ltd.,Haikou 570203,China)

机构地区：[1]海南电网有限责任公司,海口570203

出　　处：《国外电子测量技术》2020年第6期148-154,共7页Foreign Electronic Measurement Technology

摘　　要：连续域蚁群优化算法是蚁群优化算法的主要研究方向。通过分析蚁群觅食过程中的位置分布与食物来源之间的关系,提出了蚁群一类支持向量机(One-Class SVM)算法。在此算法的基础上,设计了一种电力离群用户检测算法,给出了算法的求解形式,根据高维用电负荷数据的特点,提出了一种基于改进One-Class SVM算法的电力离群用户检测方法,同时采用蚁群算法对支持向量机的训练参数进行优化,可以在样本分布不均匀、样本分布未知的环境下有效识别电力离群用户,并对其他算法的测试结果进行了比较和分析,以验证所提出算法的正确性和有效性。Continuous domain ant colony optimization algorithm is the main research direction of ant colony optimization algorithm.Based on the analysis of the relationship between the location distribution and the food source,an ant colony one class SVM algorithm is proposed.On the basis of this algorithm,an algorithm for detecting the power outlier is designed,and its solution form is given.According to the characteristics of high-dimensional power load data,an improved one class SVM is proposed At the same time,the ant colony algorithm is used to optimize the training parameters of SVM,which can effectively identify the power outliers in the environment of uneven sample distribution and unknown sample distribution.The test results of other algorithms are compared and analyzed to verify the correctness and effectiveness of the proposed algorithm.

关 键 词：蚁群算法 ONE-CLASS SVM 离群检测 电力离群 

分 类 号：TM76[电气工程—电力系统及自动化]