The Research On Identification Method Of Multiple Power Quality Disturbances

In recent years,with the extensive use of high-tech products,people's demand for power quality is getting higher and higher.On the other hand,due to the application of smart grid,new energy and other technologies,higher requirements for power quality are put forward.Therefore,the identification of power quality signal is not only the basis of the analysis,control and improvement of power supply,but also of great significance for the detection of power quality.In real life,the existence of power quality disturbance is not only a single disturbance,but also a mixed disturbance.Therefore,it is important to improve the power quality by the identification of power quality disturbance signal.In this paper,we mainly study on the identification of complex disturbance signal from two parts: feature extraction and classification.In the aspect of feature extraction,wavelet transform and S transform are used in this paper.For the S transform,this paper proposes an improved S transform with time-frequency resolution,the maximum value of the maximum amplitude of each column of the matrix of each disturbance,the minimum value of the maximum amplitude of each column,and the mean value of the mean value of the power frequency of the S matrix are extracted as three feature vector.The wavelet transform is used to extract the energy difference of each layer is obtained as a part of the feature vector.,together with the feature extraction of the improved S transform as the total feature.For the classification and identification of multiple power quality disturbance signals,the support vector machine is used as the classifier to classify the disturbance signal.Among them,the Gaussian kernel function is the key factor which affects the classification of the disturbance signal.This paper proposes a Gaussian kernel function algorithm base on amplitude adjustment and radial width adjustment,which introduct the amplitude adjustment parameters and radial width adjusting parameters to reduce the number of support vector and the computational complexity of the algorithm,and improves the average classification accuracy of multiple power quality disturbance.In order to solve the problem of parameter selection in support vector machine classifier,particle swarm optimization algorithm is used to optimize the parameters.In this paper,a kind of exponential inertia weight is proposed,and the optimal combination of parameters is obtained quickly and accurately,which is mainly about the optimization of Gaussian kernel parameter and penalty parameter,and improves the average classification accuracy of multiple power quality disturbance.The simulation results show that put the feature vector extracted by wavelet transform and the feature extraction of the S transform which improve the time-frequency resolution in theclassifier,classification recognition accuracy is improved by 3.7839% compared with the wavelet transform and S transform.The classification recognition accuracy is improved by7.5758% compared with the wavelet transform.The application of Gaussian kernel function algorithm based on amplitude adjustment and radial width adjustment improves the classification accuracy of SVM classifier and reduce the computational complexity,the number of support vectors is reduced,and the overall recognition accuracy is improved by 1.8182% compared with the original SVM.By using the particle swarm optimization with exponential inertia weight to optimize the parameters of support vector machine,the classification accuracy of multiple power quality disturbance is higher than that of the traditional particle swarm optimization algorithm,The overall recognition accuracy is improved by 0.3788%.