Classification And Recognition Of Power Quality Disturbance Signals Based On Mathematical Morphology And Deep Learning

With the rapid development of the national economy and the rapid advancement of science and technology,people have higher requirements for the power quality.At the same time,a large number of non-linear loads connected to the power system have caused many power quality disturbances.On the other hands,the application of power electronics device has brought about significant harmonics to the power system,which seriously threats the safety of consumers and equipments.Therefore,it is so significant to correctly recognize and classify the power quality disturbancs,which can improve the quality of power supply and ensure the safe and stable operation of power system.In this paper,we realized the accurate classification of power quality disturbance signals through the simulation analysis.Firstly,this paper introduces the research status of power quality problems at home and abroad,and briefly introduces the commonly used methods of analysis and classification of power quality disturbance.In addition,we establish mathematical models and Simulink physical model of power quality disturbances according to he definition of power quality and relevant standards at home and aboard,and the experiment datas obtained from the simulation are used as the basis for subsequent resrearch.Secondly,for the characteristics of simple structure and fast operation of mathematical morphology,this paper constructs a denoising algorithm based on double structural elements weighted opening-closing-closing-opening(DSEWOCCO)morphological filter,and a disturbance detection algorithm based on multi-resolution morphological gradient,which can effectively filter out the noise in the signals and locate the starting and ending time of disturbance signals,respectively.Besides,a feature extraction method based on Hilbert envelope and translation operator is proposed,which realizes the dimensionality reduction of data and retains important detail information,and the feature vectors are used as the input of classifier.In order to improve the accuracy of the classification,the dynamic inertia weight coefficient and cross mutation process are introduced to the traditional PSO,which can avoid falling into local optimum.And then optimize the parameters of the SVM classifier by improved PSO,which is in favor of a better classification of power quality disturbances.Finally,this paper applies deep learning to the power quality disturbances.This paper classifies single power quality disturbance by LSTM which has advantages in dealing with time series.And then,LSTM-DAE is proposed to classify and reconstruct the multiple power quality disturbances.first,LSTM is used to classify the multiple power quality disturbances,then select DAE with different parameters to decompose and reconstruct the multiple disturbance signals.That achieve the monitoring of a single component in a multiple disturbance signals.The experimental results prove the feasibility of the proposed method.