Data Compression And De-noising Of Power Quality Transient Signals

Electric energy is one of indispensable energy in modern society. With the development of society and the advancement of technology, power quality has been drawn extensive attention of power department and users. During the monitoring process of power quality signals, the size of the obtained data will be very large, which causes some difficulties in data storage and transmission, so the data must be compressed. On the other hand, due to external electromagnetic interference and some problems of equipment installation, power quality signals will be affected by noise that covers the real signals, which affects signal detection and restoration, so the de-noising is also very important.This thesis focuses on the problems of data compression and de-noising of power quality transient signals through applying the wavelet method and an improved MDL (Minimum Description Length) criterion. The main work and achievements are as follows:1) Since power quality transient signals are non-stationary and nonlinear, via using wavelet analysis of unique time-frequency localization, wavelet thresholding de-noising is a simple and efficient method for power quality transient signals de-noising. However, by using the traditional wavelet thresholding de-noising, Pseudo-Gibbs phenomena occur in the neighborhood of singularities. Translation invariant wavelet thresholding de-noising algorithm is introduced to denoise power quality transient signals in this paper. By piecewise shifts of the original signal, the discontinuous jump points are eliminated. The simulation results show that translation invariant wavelet de-noising method can well suppress the Pseudo-Gibbs phenomena and obtain better de-noising results for power quality transient signals in comparison to the traditional wavelet thresholding de-noising method.2) Soft threshold method and hard threshold method are the two kinds of typical threshold de-noising approaches, but both are deficient. Based on a nonlinear wavelet threshold algorithm, a recursive way to determine the optimal de-noising threshold is presented in this paper to improve de-noising effect. In practical applications where the variance of noise is unknown or changeable, this approach can estimate the noise variance and the threshold adaptively. The simulation results show the proposed approach not only improves noise estimation to a certain extent, but also the de-noising effect is better than four rules of traditional sqtwolog threshold, rigrsure threshold, heursure threshold, and minimaxi threshold.3) Since it is difficult in selecting mother wavelet and setting wavelet threshold, a novel compression and de-noising method for power quality transient signals is put forward in this paper via an improved MDL criterion which is put forward in the information theory. The novel method combines with wavelet analysis, which is used to adaptively select an optimal wavelet filter for different types of power quality transient signals and wavelet retained coefficients for signal reconstruction. In the conditions of different noise levels and types of power quality transient signals, the optimum correspondency between the conflicting goals of compression ratio and mean square error is found by using this novel algorithm. The simulation results show the method in this paper is superior to the traditional MDL.4) Finally, make a summary of the full paper, and put forward some prospects of further research.