Research On The Application Of Mathematical Morphology To Signals Analysis In Power System

On the basis of a novel nonlinear signal processing method called mathematical morphology (MM), this paper studied the power system power quality (PQ) disturbances detection and location, the partial discharge (PD) signal extracting, the leakage current of metal oxide surge arrester (MOA) extracting and the early fault diagnosis of MOA. In sum, the work in this paper is as follows:1. Based on the MM and the power algorithm of short data window, a novel method of PQ disturbances detection and location is proposed according to the characteristics of PQ signals. Firstly, in order to filter random and impulse noises, a simple and practical morphological filter based on the principle of MM is designed. It holds the fundamental characteristics of original signals when filters noises and is more efficient than wavelet and traditional morphological filters. Then, to the filtered results, the disturbances are located precisely by power algorithm of short data window according to the changing energy. The voltage swell, sag, interruption and short-duration fluctuation are used to verify its validity.2. According to the characteristics of PD, a new method using empirical mode decomposition (EMD) and MM is applied to extracting the signals from the PD signals buried in excessive noises. It dosen't need to define any basic function, so it has strong filtering adaptability. For the intensive periodic narrow bandwidth interference, a simple method combining FFT and MM is put forward to suppress the interference based on the differences between the PD and interference. The simulation and experimental results show that the EMD and MM method can efficiently remove the noises without changing the fundamental shape of the original signals. This technique solves the difficulty in denoising and preserving the features of signals. At the same time, the method combining FFT and MM can eliminate the narrow bandwidth interference, and the distortion of PD signals is very low.3. In order to extract the leakage current, a parallel-composed morphological filter with multiple structuring elements based on the principle of MM is designed. Also, pattern spectrum is used to analyse the early fault diagnosis of MOA. MATLAB simulation and experimental data analysis demonstrate the filtering algorithm can gain the real leakage current precisely and enhance the SNR (signal-to-noise ratio) greatly. Pattern spectrum gives a new ideal for early fault diagnosis of MOA.