Recognition And Classification Of Complex Power Quality Disturbances

Under the background of the era of vigorous development of green power,photovoltaic power generation,wind power generation,etc.have become larger and larger.Electric energy storage and conversion,and the use of a large number of power electronic equipment have led to ultra-high harmonic problems in power systems.And other power quality issues are becoming increasingly prominent.Through the development of high and new technology,especially the development of high-end manufacturing,higher requirements have been imposed on the quality of power supply.The evaluation and management of power problems is more and more urgent.The identification and classification of various power quality problems is an alternative to the evaluation and management of power quality problems.Therefore,the classification of power quality issues is of great significance.First,several common S transforms are introduced and analyzed from the bas ic principle,parameter selection,and time-frequency resolution of the S transform.Based on this,its existing problems are analyzed,and a new S transform is proposed,that is,an S transform based on the Blackman window and the window width ratio,and the selection of its optimal parameters is proposed,from the sampling frequency,the amplitude of the disturbance,and the duration of the disturbance.This aspect analyzes its influence on the optimal parameters,and proposes a fast algorithm based on this.Then,the difference between the time-frequency resolution of the S-transform and the traditional S-transform is analyzed through simulation data.Finally,the validity is verified by field data.Secondly,in view of the shortcomings of traditional fea ture-based feature extraction methods,such as large and complex feature value screening,two feature extraction methods are proposed.1)Based on the feature extraction method of the characteristic curve,and based on the curve energy segmentation method of wave energy density,probability mean and FFT,the noise interference is eliminated and the characteristics are completely retained,so that the characteristic curve is shortened.The method based on the eigenvalue method solves the problem of large eig envalue selection,and improves the anti-noise ability of the algorithm;2)the extraction method based on the eigenvalue matrix,extracts features from the low frequency,high frequency and third harmonic parts to reduce different disturbance Mutual interference between features,especially for spikes and notches.This lays a solid foundation for the identification and classification of complex disturbances.The matrix segmentation method is used to ensure the integrity of the information,exclude other disturbances and noise interference,and highlight the characteristics of the disturbance.Finally,a corresponding classification method is proposed for the characteristic curve and the characteristic matrix respectively.1)Aiming at the characteristic cur ve,a database query method is proposed.Simulation results show that the method based on eigenvalues avoids the large-dimensional feature selection process,and at the same time,it can achieve higher recognition and classification accuracy under high noise background.2)For the feature matrix,Alex Net’s perturbation recognition method is used to solve the problem of weak peak and notch recognition.At the same time,it can identify 96 kinds of perturbations including 6 kinds of complex perturbations.