基于小波和BP神经网络的电能扰动分类新方法

电能质量扰动的分类包括特征向量提取和分类器构建2部分。基于小波和神经网络的分类方法大部分采用小波分解各层的能量分布作为特征向量,用单个神经网络给出分类结果,此类方法构建的分类器性能有待进一步提高。文章构建了一组基于小波变换的特征向量作为分类器的输入。通过基于最小二乘法的策略综合3个相互独立神经网络的输出以得到最后的判别结果。算例表明提出的分类器准确率高,在信噪比20 dB的情况下还可以达到93.18%的准确率。分类器能有效识别电压中断、电压暂降、电压暂升、谐波、振荡暂态和闪变6种常见电能质量扰动。

A New Approach to Recognize Power Quality Disturbances Based on Wavelet Transform and BP Neural Network

Classification of power quality disturbances consists of two stages,namely,characteristic vector extraction and classifier construction.In most classification methods based on wavelet transform and neural network the energy distribution in each layer obtained by wavelet decomposition is used as the characteristic vector of the layer,and the classification results is given by neural network,however the performances of the clasifier constructed by such methods are to be further improved.In this paper,a set of wavelet transform based characteristic vectors are taken as the inputs of the classifier.By means of the strategy based on least square method the outputs of mutually independent neural networks are synthesized to achieve final recognition result.Results of calculation example show that the proposed classifier is accurate and under white noise background with the S/N ratio of 20db the obtainable reconnition accuracy is 93.18%.The proposed classifier can effectively recognize six typical power quality disturbance patterns: voltage interruption,voltage sag,voltage swell,harmonics,oscillating transient and flicker.