Research Of Power Quality Analysis Based On Sparse Decomposition

With the increasingly application of nonlinear load and power electronics,and with the continuous development of distributed power generation,high-voltage ac/dc transmission,micro network and electric vehicles in power grid,power quality problems are more serious than ever before.To this end,many researchers have worked on the relevant topic research,and gotten a lot of achievements.The massive use of new energy and the diversity of load have reduced the power quality of power grid,thus algorithms need to achieve higher accuracy of power quality feature detection and disturbance identification.Besides,the high accuracy methods will lead to a higher sampling rate and more analysis data,which reduces the realtime of power quality analyzation.Therefore,studying highly efficient power quality analyzation algorithm is very significant.Studying power quality problems mainly focuses on power quality disturbance detection method,disturbance data compression algorithm and power quality disturbance identification algorithm.To ensure high precision and high recognition accuracy,this paper,which is supported by two national natural science foundation of China,mainly studies the detection,classification and compression of power quality disturbance in terms of achieving high resolution both in time domain and frequency domain.The main research content of this article is as follows:(1)This paper proposes a power quality analyzation method based on sparse decomposition for solving the interaction problem.To ensure the high precision of the detection,classification and compression of power quality disturbance signal,it needs to reduce the interaction of its transient components and its steady-state components.For this problem,this paper studies different kinds of dictionary base which is suitable for analyzing power quality signal and introduces analytical abilities of these bases in the time domain,frequency domain and time-frequency domain.Besides,the evaluation standard of the correlation between them is provided,the standard is based on cross correlation and Pearson correlation coefficient.By the standard it has been proved that the combination of unit matrix and Fourier basis is the optimal joint-domain dictionary.In addition,this paper analyzes several greedy algorithm for obtaining sparse coefficients,then summarizes the applications and the improvement direction of them in the field of power quality analysis.(2)In order to avoid the additional storage space of the imaginary part of Fourier basis,a joint-domain dictionary based on unit matrix and Hartley basis is proposed.Hartley basis is the matrix form of Hartley transform coefficients.As Hartley transform is a linear transformation of Fourier transform,Hartley basis,which has the same analytical ability in frequency domain with Fourier basis,approximates the optimal joint-domain dictionary.By employing this dictionary,a Joint-domain dictionary mapping(JDM)parameter estimation algorithm of power quality disturbance is proposed.This method can separately represent transient component and steady-state composition,and the mapping results of the transient and steady-state components are the time domain waveform and the frequency spectrum,respectively,thus it can accurately detect the characteristic parameters of different disturbance components.The comparative experiments of related algorithms show that the algorithm has higher accuracy and better resistance to noise.Moveover,JDM algorithm is verified by the experiments of real data on the practicability and reliability in actual situation.(3)With the growing number of power quality disturbance monitoring,the amount power quality disturbance waveform(PQDW)data also increases,thus it needs to effectively compress the large monitoring data.This paper proposes a disturbance component compression method based on JDM algorithm.Due to the fact that joint-domain dictionary can express the transient and steady-state components independently and output compactly supported coefficient of each component,the compression effect of the proposed algorithm is better than those of related algorithm.To quickly complete a dictionary mapping,this paper proposes a Joint-domain orthogonal matching pursuit(JOMP)algorithm.JOMP,which locates transients in time domain and steady-state components in frequency domain by wavelet transform and Fourier transform,respectively,significantly improve the efficiency of the atom selection and guarantee the accuracy of data reconstruction.When high compression ratio is required,JDM compression algorithm can ensure the accuracy of disturbance data characteristics better compared to other algorithms in the process of compression.The experiments of real data and field monitoring data show that the proposed algorithm has a good compression effect and application prospect.(4)Propose a highly accurate and fast power quality disturbances classification using dictionary learning sparse decomposition(DLSD).Firstly,an over-complete dictionary is constructed by combining an identity matrix with a learning dictionary trained by K-SVD algorithm.Secondly,the features and the fuzzy primary classifications of power quality disturbances are obtained by calculating the sparse decomposition coefficients based on the learning dictionary.Then a decision tree is adopted to accomplish accurate classification by using the estimated features and the pre-classification results.For being adaptive to sparsity and reducing computational complexity,a fast adaptive matching pursuit using sparsity adaptive algorithm and regularized atom selection is proposed.Finally,the proposed approach is tested by power quality disturbances from simulations and actual measurements.