基于最优小波包基分解的暂态电能质量分类方法

Transient power quality classification based on best wavelet packet basis decomposition

对电能质量暂态扰动进行正确的识别分类是改善电能质量的前提,而电能质量扰动特征向量的提取又是电能质量扰动识别分类中的关键步骤。提出基于最优小波包熵特征的特征提取方法,对采样信号进行小波包分解及时域预处理并选取最优小波包基,计算各尺度下信号的最佳小波包子空间的熵值,归一化处理后,把同尺度下的熵值和作为特征量,再将所有尺度下的特征量按尺度分解顺序依次组合在一起,形成最终的特征向量并作为神经网络的输入构建神经网络识别系统,对暂态电能质量信号进行识别。系统负荷投切和电容器充电的仿真结果表明,该方法能快速有效地区分暂态脉冲和振荡暂态。

Correct identification and classification of power quality transient disturbance is the precondition to improve power quality,while the eigenvector extraction of power quality disturbance is the key step of the classification. A way of extraction based on the entropy the best wavelet packet is presented. After wavelet packet decomposition,time-domain preprocessing and wavelet packet basis selection for the sampling signals,the entropies of best wavelet packet subspace in different scales are calculated and normalized. By taking the sum of entropies in same scale as an eigenvalue and combining the eigenvalues in all scales according to their decomposition sequence, the final eigenvector is thus formed and then input to an NN（Neural Network ） to construct a NN identification system for transient power quality. The simulations of load switching and capacitor charging show that the proposed method can quickly and effectively class the impulse transient and oscillation transient disturbances.