电能质量扰动识别的小波压缩感知方法

为改善基于小波电能质量信号扰动识别中存在数据量大、识别率不高的不足,提出一种电能质量扰动识别的小波压缩感知新方法。该方法首先确定扰动信号在小波域中的稀疏性,利用小波压缩感知降维,获得少量测量数据,应用正交匹配追踪算法求取各层稀疏系数组成稀疏矩阵;然后提取稀疏系数的最大值、标准差、峭度等组成特征向量,输入神经网络系统训练并实现分类识别。该方法具有采样数据少、处理方便、特征提取简单等特点。仿真结果表明,针对典型的7类单一扰动和复合扰动信号,所提方法在理想环境下识别率分别达到99.50%和99.43%,噪声环境下识别率分别达到97%和98%以上,拥有较强的鲁棒性和较好的准确性。

Power Quality Disturbance Recognition Method Based on Wavelet Compressive Sensing

To overcome the disadvantages in the power quality signal disturbance recognition based on wavelet transform,such as a large amount of data and a low accuracy rate,a novel power quality disturbance recognition method based on wavelet compressive sensing is proposed.First,the sparsity of disturbance signal in wavelet domain is determined.Second,wavelet compressive sensing is used to reduce dimensions,and a few measurement data can be obtained.Then,the sparsity coefficient in every layer can be obtained by orthogonal matching pursuit(OMP)algorithm to form a sparsity matrix.Finally,eigenvectors are formed by the extracted maximum,standard deviation,kurtosis,etc.,and they are further input into the neural network system for training so as to classify and identify the types of disturbance signal.This method is featured by less sampling data,easy process,and simple extraction of characteristics,etc.Simulation results demonstrate that the recognition rates obtained using the proposed method for seven typical single and mixed disturbance signals reach 99.50%and 99.43%respectively in an ideal environment,and above 97%and 98%respectively in a noisy environment,showing stronger robustness and better accuracy.