As the power quality disturbances analysis based on traditional generalized S-transform (GST) is of large amount of calculation,it is not conductive to the real time detection and classification of signals.Moreover,and there exist problems including low precision detection and low classification accuracy when the signal-to-noise ratio is not higher enough.In this paper,the traditional GST algorithm is improved and applied to the analysis of power quality disturbances.Firstly,fast Fourier transform is used to estimate signal frequency,narrow the range of frequency domain analysis,and saves the computation time obviously.Secondly,considering that the disturbance amplitude curve changes slowly at the start and stop time when using the traditional S transform,bi-gaussian window is used instead of traditional Gaussian window,and an adaptive parameter selection window function is used to obtain the analog matrix information more reliably after transformation.Finally,with the aid of MatlabR2010b simulation platform,a new index is introduced to accurately estimate the start and stop time information of disturbances,extract effective features to conduct classification directly,or identify particular disturbances by means of simple discrimination tree,which improves the efficiency and accuracy of classification.Through the analysis results of 12 kinds of power quality disturbance signals,the effectiveness of the proposed method is verified.%基于传统广义S变换的电能质量扰动分析方法计算量大,不利于信号的实时检测与分类,且存在信噪比不高时检测精度仍较低和分类正确率不高的问题.该文对传统广义S变换算法进行改进并应用于电能质量扰动分析.首先,利用快速傅里叶变换估计信号频率,缩小频域分析范围,大幅度节省计算时间;其次,用双高斯窗替代传统高斯窗,解决传统广义S变换检测扰动起止时间的幅值曲线变化缓慢的问题,并通过自适应选择双高斯窗参数,信号变换后得到的模时频矩阵信息更加可靠.最后,借助MatlabR2010b仿真平台引入新的电能质量扰动指标准确估计扰动起止时间信息和依据提取有效特征信息直接分类或借助简单的判别树识别特定扰动,提高了分类效率和正确率.通过对12种电能质量扰动信号的分析结果,验证了文中方法的有效性.
展开▼
