The Analysis Of Power Quality Disturbances Based On S-transform

In recent years, with the increase of nonlinear impact load in grid and frequency control devices, it results power quality problems such as grid interruption, voltage fluctuations,voltage sags, and harmonics in voltage distribution have become more and more serious. In order to suppress and control power quality problems, the device must be added to compensate for the power quality in the grid, development and tuning of these devices require accurate, detailed power quality parameters. Therefore, understanding the mechanism of the disturbance signal, accurate detection of the characteristic parameters of signal disturbance(including amplitude, frequency, start and end time, phase, etc.), and the disturbance signal classification play a very important role in improving the quality of power supply and ensuring the safe operation of the power system.This paper with Matlab platform builds five single disturbance signals(sinusoidal voltage signal, voltage sags, voltage swells, voltage interruptions, harmonics) and two complex disturbance signals(harmonic dips, harmonics temporarily liters). By time-frequency decomposition of the above signals and analysis of correlation characteristics in S-transform curve of each disturbance signal,this research proceeds each disturbance signal positioning, amplitude detection, frequency component detection, and phase detection, and analyzes the advantages of correlation curves which is proposed by this paper in detecting disturbance signal information. Compared with analysis of ignoring the phase information of signal disturbance in the past, another obvious advantage of this paper is to analyze the changes in the phase of the occurring signal disturbance, thus providing a parameter basis for the latter grid power quality compensation equipment. By contrast with the short time Fourier transform, the results show S transform detection has high accuracy, a better time-frequency resolution, and high noise immunity.Besides the disturbance signal detection,, this paper completed the automatic identification of these seven kinds of signals based on RVM classifier. Firstly, the feature vector is extracted by S-transform of the disturbance signal, and then the select eigenvectors to conduct RVM training and recognition. This paper extracts six kinds of feature vectors, and gives the comparison of recognition results under different SNR conditions. Simulation results show that this method for disturbance signal type identification has a high accuracy. Since only a few feature vectors are chosen, the whole process is less time-consuming.