Study On Power Quality Disturbances Compression And Reconstruction Method Based On Matching Time-frequency Atoms Framework

In recent years,The power department and users have paid close attention to theincreasingly prominent qower quality problem. The key of improving power quality isthe ability to analyze large amount data obtained from real-time monitoring of the grid,also effectively detect and locate、compress and store、classify and identify thedisturbances. In allusion to the problems that power quality disturbance signals possessa lot of localized information and it is difficult to extract effective and fine features ofthe signals flexibly and sententiously via the traditional methods, a framework matchingtime-frequency atoms,as well as its improved methos,is proposed to decompose andreconstruct power quality disturbances as well as extract feature parameters ofdisturbances. The main research contents and results of this paper are as following;First and formost, A novel approach matching Gabor time-frequency Atom ispresented to decompose and extrat the wispy features’ parameters of power qualitydisturbances, as well as reconstruct and compress the disturbances data. Based on thediscrete optimization of redundant Gabor dictionary of time-frequency atoms, thedisturbance signal is adaptively decomposed by matching pursuit algorithm, and theiteration times or the residual error threshold are set as the termination condition toattain a series of time-frequency atoms and their parametrization forms. In particular,the time-frequency atoms parameters and atoms are coded and encoded to complete thetasks of Power Quality Disturbances Compression And Reconstruction. The simulationresults show that this method can reconstruct and compress the power quality transientdisturbances concisely and flexibly with higher SNR(up to50dB), nice ERP (more than0.99) as well as lower MSE(0.001levels), and this method has more multi-resolutionanalysis ability to satisfy the need of power quality analysis, compared to the waveletpacket transform method.Secondly,An improved algorithm O-MP,based on orthogonal optimization oftime-frequency atom decomposition is proposed to improve the algorithm convergenceand sparse decomposition accuracy, Particularly, the Schimidt orthogonalization isapplied to the optimal time-frequency atoms being sought to update the projection spaceand reduce the redundant components, and thus to complete the tasks of power qualitydisturbances compression and reconstruction, as well as features’ parametersextraction. The simulation results show using the O-MP, the algorithm convergence and sparse decomposition accuracy are further improved relative to the sparsedecomposition based on matching pursuit.Thirdly,An improved algorithm GA-MP, using time-frequency atom decompositionoptimized by Genetic algorithm is proposed to reduce computational complexity rateand matching time. It is based on the matching time-frequency Atoms framework. Inparticular, the time-frequency atoms parameters are optimized by Genetic Algorithm toreduce the complex rate of finding the best atoms. And thus some best time-frequencyatoms matching features of disturbance signals, and also its reconstruction parameters,are obtained to complete the tasks of power quality disturbances compression andreconstruction. The simulation results show using GA-MP the Computationalcomplexity reduction rate is up to95.8%, thus the efficiency of matched disturbancefeature (up to80-100times) and convergence performance are improved further to meetthe demand of power quality analysis.Last but not least, An improved algorithm GA-OMP, bases on O-MP and GA-MPis proposed to improve the algorithm convergence and sparse decomposition accuracyas well as reduce computational complexity rate and matching time. In particular, thetime-frequency atoms parameters are optimized by Genetic Algorithm to reduce thecomplex rate of finding the best atoms. and these best time-frequency atoms areoptimized by Schimidt Orthogonalization to update the projection space and reduce theResidual component, then.the time-frequency atoms parameters and atoms are codedand encoded to complete the tasks of power quality dsturbances compression andreconstruction. The simulation results show using GA-OMP, the compressionreconstruction performance,as well as algorithm convergence are improved relative toMP and GA-MP, on the basic of keeping the advantages of computational complexityrate and matching time in GA-MP to further satisfy the need of power quality analysis.