The Research On Identification Method Of Multiple Power Quality Disturbances

With the continuous deepening of smart grid construction,the composition of power sources and loads has undergone major changes.On the one hand,the use of new types of electrical equipment,the large-scale grid connection of new energy sources,and the large number of non-linear,impact,and fluctuating loads have seriously affected the power quality.On the other hand,the use of more and more sophisticated electronic devices places puts forward higher demands on power quality.Power quality disturbance identification is the key to monitoring and improving power quality,and it is also a key indicator for evaluating the quality of power.In the actual power system,there is not only a single disturbance signal,but also a multiple disturbance in which the single disturbances are added together.Therefore,the recognition of the multiple disturbance of power quality is very important.This paper focuses on the recognition of multiple disturbance signals from two aspects: feature extraction and classification identification.In the aspect of feature extraction,the feature extraction method based on improved S transform and wavelet energy distribution is proposed.This method combines the multi-scale analysis of wavelet transform with the flexible resolution of the improved S-transform.The S transform is deeply studied and the S-transform with a flexible resolution is proposed.Using the improved S transform to analyze the time-frequency information of the disturbance signal,and extracted the maximum,minimum,mean,standard deviation,the normalized mean,skewness,and kurtosis from the improved S transform modulus time-frequency matrix as part of the feature vector.The wavelet transform is used to extract the energy difference between the disturbance signal and the standard signal.The energy differences are used as another part of the feature vector,and combined with the features extracted by the improved S transform as the total feature vector.In the aspect of classification recognition,the support vector machine is used as classifier in this paper.There is no clear and effective standard for parameter selection in support vector machine.The method determines the optimal parameters by using particle swarm optimization algorithm,and ultimately improves the recognition accuracy.In order to improve the classification performance of support vector machine,a support vector machine classifier based on hybrid kernel function is proposed.The use of hybrid kernel function instead of the original kernel function not only ensures the learning performance of the classifier but also takes into account the generalization performance,and the method improves the identification accuracy of the disturbance signal.The results of simulation experiments show that the identification accuracy of feature extraction using improved S transform is 2.197% higher than that of feature extraction using S transform.The identification accuracy of extracted features using wavelet transform and improved S transform is better than that of improved S transform increased by 5.303%,the identification accuracy of S transform increased by 7.5%;the whole identification accuracy of the SVM optimized by the particle swarm optimization algorithm is better than that of unoptimized SVM by 0.9091%;The hybrid kernel function support vector machine classification method considers both learning performance and generalization ability and the identification accuracy is improved compared to single kernel function.The identification accuracy of the radial basis kernel function is improved by 2.1213%.The identification accuracy of the polynomial kernel function is improved by 3.0303%.