Research On Power Quality Disturbance Detection And Compressed Sensing Data Reconstruction Based On Sparse Representation

With the development of science and technology,and the progress of society,the related fields of power industry have gained considerable development.The traditional power generations such as thermal,hydro and nuclear increase steadily,the new energy power generations such as wind,photovoltaic,biomass,geothermal are booming.The conversion equipment of distributed generations and various intelligent power supply equipments emerge in endlessly,and meanwhile the power supply and consumption increase with each passing day,these entire make the research field of electric power science and technology expand ceaselessly especially with the conception of smart grid.The power quality has become a hot research field in recent years.The research in this field can provide important support for safe and high quality power supply and the economical operation and sustainable development of the modern power system,and also has important theoretical research value and great economic value and social significance.Based on the sparse representation theory,this paper focuses on the research of disturbances detection and compressed sensing reconstruction of power quality signal.The main research works are summarized as follows:The sparse representation is applied to the power quality signal processing,and a fast decomposition method of matching pursuit based on over-complete dictionary is proposed.The convergence of the matching pursuit algorithm and the time-frequency distribution of the matching atoms are demonstrated,and the mathematical models based on optimization algorithms to search the most matched atoms are established for the over-complete dictionaries of discrete and continuous parameters respectively.These provide a reliable theoretical basis for the study of the sparse decomposition of power quality signal and the detection and analysis of disturbance parameters.On the basis of deeply researching on the characteristics of power quality disturbance,the matching pursuit sparse decomposition method in over-complete dictionary is improved,the detection of the disturbance parameters and the extraction of the waveform of the disturbance feature are realized.This method combines the global search and local optimization algorithm to obtain the parameters of the most matched atom in the Gabor dictionary.By utilizing the time-frequency aggregation of Gabor atoms,the computational complexity of the global search is reduced by the calculation of non-zero inner product,and then the initial values of the disturbance parameters are obtained quickly.Further,the optimal Gabor atomic parameters matching the disturbance are obtained by using the unconstrained optimization algorithm,and the starting and ending time of the disturbance are determined by using the inner product recursive calculation method of atom and residual.The synthetic dictionary which conforms to the characteristics of power quality disturbance is constructed,and the disturbance parameters are used to extract the characteristic waveform of disturbance.The effectiveness of the proposed method is verified by simulation analysis and actual signal testing.Aiming at the difficulty of estimating the sparsity level of the transformation coefficients of power quality signal,the problem of the compressed sensing reconstruction of power quality signal is studied in the framework of greedy algorithm.According to the disturbance characteristics of power quality,the selection of the compressed sensing matrix and sparse transform domain are discussed.Based on the existing SP and SAMP algorithms,a new sparsity adaptive ASMP algorithm is proposed.This algorithm expands the support set by step size without prior estimation of the sparsity level.The simulation analysis proves that the reconstruction performance of the proposed algorithm is better than the existing algorithms.Considering the difficulty of obtaining the sparse transform domain of the complex power quality disturbance signal,the performance of compressed sensing reconstruction is improved by using the block sparsity of the transformation coefficients.The block sparsity of the power quality signal with complex disturbances in different transform domains are analyzed,and the BSBL_BO algorithm based on block sparse Bayesian learning framework is adopted.When the sparsity of the transformation coefficients is poor,the prior distribution and posterior probability information and the learning rules for distributed parameters are used to adaptively learn and exploit the intra-block correlation and thus reduce the requirement for sparse transformation,so that the signal reconstruction quality is further improved.