Detection And Classification Of Power Quality Disturbance Based On VMD And ELM

With the accelerating process of social informationization and industrial automation,all kinds of nonlinear and impact loads are connected to modern power grid,which also brings a lot of power quality problems.Accurate detection of disturbance parameter information and identification of various disturbances are the premise and key to solve power quality problems.This thesis analyzes and improves the problems existing in the mainstream application methods in the field of disturbance detection and classification,and the main contents are as follows:(1)The principle,advantages and disadvantages of Hilbert-Huang Transform(HHT)are briefly introduced in this thesis.For the mode mixing of Empirical Mode Decomposition(EMD)in HHT,this thesis adopted the method called Adaptive Variational Mode Decompoisition.Since the artificial selection of the mode value k in the VMD algorithm will directly affect its decomposition effect,this thesis calculates the coefficient of multiple correlation between the disturbance signal to be measured and each IMF component,so as to complete the selection of the mode value k adaptively.Aiming at the endpoint effect in EMD and VMD,the mirror extending method is used to process the original signal.Experimental results show that the improved VMD can effectively improve the problems of mode mixing and endpoint effect in HHT.(2)In view of the power quality disturbance detection,this thesis has made a deep study in terms of transient,steady and composite disturbance,combined the improved VMD method with Hilbert transform and then applied them to the detection of disturbance signals.First of all,the modal value k of VMD is determined by using the coefficient of multiple correlation,and then the extended disturbance signals can be decomposed by adaptive VMD.Next,the instantaneous amplitude frequency information corresponding to each IMF classification is obtained by HT transform.Finally,the mean fitting method and difference calculation method are used to accurately locate the starting and ending time of disturbance.Simulation results show that compared with EMD and EEMD,the method of adaptive VMD combined with HT transform has higher detection accuracy for various power quality disturbances.(3)The combination of VMD and the optimized ELM is applied to the power quality disturbance classification.Firstly,the VMD method is used to process the power quality disturbance signal,to calculate the energy value of each disturbance component and then normalize it,and taking the processed energy value as the eigenvector of the disturbance signal.Secondly,by selecting ELM as the classifier of disturbance signal,the usage of Grey Wolf Optimizer(GWO),which has a strong balance between the local optimization and the global search is in favour of its optimization.Finally,the normalized eigenvalue is used as the input vector of GWO-ELM,to realize the classification and recognition of various disturbance signals.Experimental results show that GWO-ELM has higher recognition rate for single and composite disturbance than GA-SVM and GA-ELM in noisy environments with different signal-to-noise ratios.The thesis has 62 figures,20 tables,81 references.