Research On Sparse Representation And Compressive Sampling Of Power Quality Disturbance Signals

With the scale of power increasing rapidly, a variety of power electronic devices and non-linear load are applied in the power grid, making the harm of power quality in power system serious. The problem of power quality not only catches wide attention by related areas, but also becomes the urgent problem that power companies and users want to solve.So the effective real-time monitoring and analysis for power quality is very important. In order to be able to analyze the power quality data effectively, the power quality data should be sampled and compressed. However, existing monitoring systems are based on the Nyquist theorem,its massive sampling data brings not only enormous pressure to the acquisition devices,but also great difficulty in data storage and transmission. The appearance of compressed sensing theory breaks the bondage of the Nyquist theorem with regard to the sparse or compressible singals, and makes them can be sampled and compressed simultaneously. The theory can reconstruct the original signals through a small number of samples,which not only improves the sampling and transmission rate but also reduces the storage space.This paper mainly studies the two basic problems of compressed sensing:the sparse representation of power quality signal and the design of measurement matrix. The main achievements are summarized as follows:(1) According to the definitions and standards of power quality, we modeled a series of common power quality signals by MATLAB, including the singal interference model and multiple interference model. From a practical point of consideration, we built IEEE14 nodes system by PSCAD to obtain three-phase line fault signals.(2) According to the characteristics of the harmonics’sparse representation in the Fourier domain,we propose a harmonics detection method which is based on OMP. The new method can directly detect the interested base wave and harmonic components through the reconstructed data within a certain degree and extent.(3) The main purpose of compressed sensing is the sampling of power quality data effectively.We improve two kinds of measurement matrices which are constructed simply based on the principle of structured measurement matrix. These matrices occupy less storage space and are deterministic. Furthermore, the two kinds of matrices are better than(or are equivalent to) common measurement matrices on reconstruction accuracy and more friendly to hardware design.(4) Because pulse, transient oscillation,some multiple disturbance and three phase fault signals can’t be reconstrucyed accurately by using Fourier sparse basis, we propose a compression sampling method based on discrete stationary wavelet transform. Simulations and experiments show that this method can accurately reconstruct these signals. Furthermore,the method is more general. The good performance meets the requirements of power quality analysis largely.