基于GA和ELM的电能质量扰动识别特征选择方法

电力系统中海量暂态扰动的分析与治理需要以高效准确的扰动分类为基础。现有扰动识别方法缺少合理的特征选择环节,分类器过于复杂,不能满足高效分类的需要。提出一种新的电能质量扰动特征选择方法。首先,对原始信号使用S变换进行预处理,提取具有代表性的25种扰动信号特征构建原始特征集合;然后,根据极限学习机识别准确率构造用于扰动特征选择的遗传算法适应度函数;最后,用遗传算法来进行迭代运算,确定最优特征集合。实验证明,新方法能够有效去除冗余特征,在保证分类准确率前提下,有效降低分类器复杂度,提高分类效率。

A Method of Power Quality Disturbances Recognition Feature Selection Based on GA and ELM

Analysis and management of massive transient disturbance in power system need to take high efficiency and accuracy of disturbance classification into consideration .Since existing disturbance identification methods are lack of rational feature selection process and the classifier is too complicated to meet the needs of efficient classification .In this paper, a new method with feature selection of power quality disturbance is proposed .Firstly, the original signal is preprocessed by S-transform and 25 kinds of disturbance signal features are extracted which are representative to build the original feature set .Secondly , based on extreme learning machine recognition accuracy rate , the fitness function of genetic algorithm used for the disturbance feature selection is built .And finally, using genetic algorithm is used to conduct iterative arithmetic and determine the optimal set of features .Experiments show that the new method can ef-fectively remove redundant features , reduce classifier complexity and improve efficiency classification with ensuring the accuracy of classification .