基于优化最小二乘支持向量机的电能质量扰动分类

Classification of Power Quality Disturbances Based on Optimized Least Squares Support Vector Machine

提出了一种基于最小二乘支持向量机(Least Squares Support Vector Machine,LS-SVM)分类器的电能质量扰动分类方法,对电网环境中多类扰动特征混合的情况进行更加精细的分类辨识。针对电能质量扰动特征向量的特点,对混合蛙跳算法(Shuffled Frog Leaping Algorithm,SFLA)进行了改进,与交叉验证(Cross Validation,CV)相结合实现了对多分类器参数的优化,有效地解决了支持向量机模型参数优化的问题。仿真实验与工程验证表明,经过优化后的分类器不仅可以提高电能质量波形的分类精度,还可以进一步使分类器避免出现过学习的状态,有效提高了分类器的计算速度。

This paper presents a classification method of power quality disturbances based on least squares support vector machine（LS-SVM）. Through this method, the mixed power quality disturbances in grid can be classified and identified in detail. Further more, for the characteristics of the power quality disturbances feature vector using improved shuffled frog leaping algorithm（SFLA） and cross validation to achieve the optimal classifier. It effectively solves the SVM model optimization problem. Simulation and engineering results show that the optimized classifiers not only output high classification accuracy in small training set case, but also improve the classification performance further and effectively avoid the state of excessive learning.