一种电能质量多扰动分类中特征组合优化方法

针对电能质量扰动分类中冗余特征量造成分类器训练困难、分类准确率下降的问题,提出一种基于改进遗传算法的特征组合优化方法。该方法对信号进行小波变换,提取各层的改进小波能量熵作为原始特征,并构造一种基于欧氏距离的适应度函数,采用改进的自适应遗传算法对原始特征进行筛选和优化组合,形成用于电能质量扰动分类的最优特征组合。分别采用二分类-支持向量机法(Binary-SVM)、多标签径向基神经网络(ML-RBF)和多标签K近邻法(ML-KNN)对不同噪声情况下的电能质量单一扰动和混合扰动进行分类,仿真结果验证了所提特征组合优化方法能有效提高分类器的训练速度和分类准确率。

Feature combination optimization for multi-disturbance classification of power quality

Aiming at the difficult classifier training and low classification accuracy of power quality disturbance classification due to the redundant feature parameters, an optimization method based on the improved genetic algorithm is proposed for the feature combination, which carries out the wavelet transform to extract the improved wavelet energy entropy as the primal feature for each layer of signals, constructs a fitness function based on Euclidean distance, applies the improved adaptive genetic algorithm to optimally select and com-bine the primal features as the optimal feature combinations for the power-quality disturbance classification. Single and mixed power-quality disturbances in different noise conditions are classified by three multi-label classifiers(Binary-SVM, ML-RBF and ML-KNN) respectively, and the simulative results show that, the proposed optimization method of feature combination effectively improves the training speed and classification accuracy of different classifiers.