Research On Power Quality Disturbance Detection And Classification Method

With the development of industrial technology and the use of precision equipment,power users are increasingly demanding power quality.Unqualified power quality has caused huge economic losses.In order to ensure the user’s requirements for power quality,the management of power quality cannot be delayed,and power quality disturbance detection and disturbance identification are the premise of power quality management.Therefore,this paper studies the two aspects of power quality detection and classification,and provides a basis for the governance and protection of power quality disturbance.Aiming at the problem of power quality disturbance detection,a new nonlinear and non-stationary signal analysis method--Variational Mode Decomposition,is introduced.A method of power quality disturbance analysis combined with variational mode decomposition and Hilbert transform is proposed.The method uses the variational mode decomposition algorithm to decompose the power quality disturbance signal to obtain several finite bandwidth natural modal components.Then,the Hilbert transform is performed on each modal component to obtain the instantaneous amplitude and instantaneous signal.Frequency,detecting the instantaneous frequency and amplitude to obtain the start and end time and amplitude change of the disturbance.Aiming at the classification method of power quality disturbance in traditional machine learning,the characteristics of power quality disturbance based on discrete wavelet transform are studied.In order to further improve the effect of power quality disturbance classification,feature selection algorithm is adopted to select the best features.A new multi-sensor perceptron extreme learning machine for power quality disturbance classification model is studied and analyzed.Compared with the traditional classification algorithm,the model introduces a module of multi-layer perceptron,which can improve the training speed and classification speed of the model.The proposed model has high accuracy in power quality disturbance recognition,and its performance is relatively stable.Compared with traditional machine learning algorithms,it has better classification effect.In order to quickly and efficiently process a large number of power quality disturbance data,and at the same time solve the shortcomings of slow data acquisition and transmission,and further improve the efficiency of power quality disturbance recognition,a pattern recognition method based on compressed sensing data and deep learning algorithm for power quality disturbance signals is proposed.First,the use of compressed sensing technology to collect data can reduce the sampling frequency and reduce the storage space of the data acquisition device and increase the transmission efficiency,enabling lightweight data collection;Secondly,the data collected by the compressed sensing technology is used as the input data of the convolutional neural network to realize the automatic feature extraction and closed-loop feedback of power quality disturbance,which is completely dependent on data driving.In view of the shortage of measured signals,the amount of data of individual disturbance types is small,and the data is unbalanced,the data is enhanced to process the measured data and the simulation model is trained.