Research For Detection And Identification Of Power Quality Disturbances Based On FRGST And Combined Classifier

With the wide application of impact,nonlinear and sensitive loads,and the increasing of power system structure complexity,in addition to the traditional power quality problems,other power quality problems,such as inter-harmonics,voltage sag and transient oscillation,produce a huge threat to power grid security.Therefore,it is of great theoretical and practical significance to seek a comprehensive and accurate method of power quality to realize the detection and analysis of power quality parameters.In this thesis,the problems of power quality detection and identification are studied.At present,there are still some problems in the detection of non-stationary power quality signals,such as the lack of accuracy and the low resolution of inter-harmonics detection.Combining generalized S transform and fractional Fourier transform(FRFT),a power quality disturbance detection method based on fractional generalized S transform(FRGST)algorithm is proposed in the thesis.First of all,performing FRFT algorithm for power quality disturbance at each order.Then obtaining the optimal order by searching for spectral peak.Finally,performing FRGST algorithm at the optimal order to get the amplitude,frequency and duration of the disturbance.The algorithm is verified by the single and multiple power quality disturbances in MATLAB.The simulation results show that FRGST algorithm can realize amplitude,frequency and duration detection of the single steady-state disturbances,single transient disturbances and multiple disturbances.Compared to generalized S transform,FRGST has higher time-frequency focusing performance and detection accuracy.In order to solve the problems that the difficulty of features extraction and the low accuracy of classification for multiple power quality disturbances,a method combined decision tree and SVM optimized by quantum particle swarm(QPSO-SVM)is proposed.First,extracting six features of the signals using FRGST,and constituting a feature vector.Then optimizing the kernel function and regularization parameter of SVM by QPSO algorithm.Finally,realizing high accurate classification while the decision tree as main classifier,SVM as an auxiliary classifier.Constructing 8 kinds of single disturbances,9 kinds of double disturbances and 6 kinds of triple disturbances in MATLAB,and implementing classification using the proposed method.The simulation results show that the classification method proposed in this thesis has higher classification accuracy,stronger noise resistance,and better stability compared with decision tree and multi-label RBF(ML-RBF),so that it can meet the needs of practical application.In this thesis,an experiment platform based on Lab VIEW is built.Realizing detection and analysis of power quality data measured in the second rolling workshop of Shanxi Xishan Coal Dehui Industrial Co.Ltd.,respectively using HIOKI 3196 power quality analyzer and Lab VIEW platform.At the same time,two transient power quality disturbances,include voltage sag and transient pulse,are used to test the platform.The practicability of the platform,as well as the feasibility and accuracy of the proposed algorithm,is verified by experiments.