Power Quality Disturbances Detection And Identification Based On Signal Processing Theory

With the change of loads in power systems, the power quality problems are becoming increasingly prominent. Power quality problems not only endanger power system security and stability, but also bring vast economic loss. Based on the previous research, several key issues, such as power quality disturbance signal denoising, power system disturbance detection and identification are comprehensively and thoroughly studied and analyzed in this thesis.Denoising based on wavelet transform and mathematical morphology have been proven to be effective, but whether they can work well when used in power disturbance signal denoising and how to choose the algorithm according to different signals and different accuracy requirements become a very important issue. A remodeling factor was established for evaluating the ability of algorithm in reconstructing signal. Through analyzing the simulation results of several typical power disturbance signals, calculated the remodeling factor then discussed the adaptability of different denosing algorithms used in different signals'denosing.. The result of the computer simulation indicates that for the signals without pulse or high frequency oscillation, disturbance both methods are effective, which should be chosen according to the evaluating speed and the different accuracy requirements.For the signal with pulse or high frequency oscillation disturbance, the denoising based on wavelet transform is Obviously better than the denoising based on mathematical morphology.Because of the good effect on preserving abrupt signal, mathematical morphology(MM) often be used in detecting and locating short-time Power Quality Disturbance, but some methods based on mathematical morphology still has the shortcoming that some disturbances which crosses the zero spot couldn't be detected.. So this article has analyzed three method for detecting and locating power quality disturbance which based on Mathematical morphology, there are the method based on derivation and morphology gradient (MG),the method based on morphology gradient and soft threshold and the method based on Dq analysis and Top-hat transform. Compared adaptability of each method on the analysis of some common Power Quality Disturbances by emulating, finally found the method based on Dq analysis and Top-hat transform has good effect on detecting the disturbance which crosses the zero spot, so chose this method, analyzed some actual data. The result indicated that this method can detect and locate disturbances occurred at any time, proved it has good adaptability and feasibility.Cohen distributions are joint time-frequency distributions in a real sense, learning the basic theory, a discussion was taken to study the possibility of power quality disturbance identification algorithm based on Cohen Bilinear Time-Frequency distributions in this thesis. A new method based on rearranged bilinear Time-Frequency distribution is introduced for power quality disturbance detection. First, separate fundamental component and disturbance with instantaneous reactive power theory and generalized morphological filter; then analyze the disturbance with arranged bilinear Time-Frequency distribution, accordingly for better time-frequency concentration. Simulation results of different PQ disturbances and crossed PQ disturbances have proven the effectiveness of the proposed method on the display of the time-frequency characteristics, but the quantitative feature extraction and the cross-term suppression problem are still unresolved.Proposed a PQ disturbances classification method based linear time-frequency distribution (windowed Fourier transform and S-transform) and binary threshold feature matrix. Combined the advantages of WFD and ST, the method presents five features and binaries them, constitutes a binary threshold feature matrix, classifies different disturbances through comparing the magnitudes of the binary feature to the binary threshold feature matrix. Simulation results of 9 common kinds of disturbances indicate that the method has good performance of accuracy(>98%) and shows the validity and efficiency of the method.A layered analysis systems was bulid for power quality identification, new features and algorithms were proposed to complete the function of the system which contains seven function modules. Using dq transform, generalized morphological filter, Fourier transform and others simple signal processing methods, these modules extracted amplitude disturbance time, disturbance frequency domain singular entropy and other features and classified them layer-by-layer, then comprehensive judgment was made for classification results of every layer. Simulation results of 6 common kinds of single disturbances and some mixed disturbances indicate that the method is effective.