キャンパスWAMSによる改良されたヒルベルトホーン変換を用いた電力系統動揺特性解析

摘要

This dissertation presents a complete oscillation monitoring system based on real-time wide-area measurements from PMUs. This oscillation monitoring system employs the enhanced Hilbert-Huang transform (HHT) to analyze power system oscillation characteristics and estimate the damping of oscillatory modes from ambient data. This new oscillation system can give an indication of the damping of transient oscillations that will follow a disturbance, once it occurs. The application is based on a system identification procedure that is carried out in real-time.This research studies various low frequency oscillation analysis algorithms. It mainly introduces the concept, character and implementation process of FFT, WLT and HHT method. According to the characteristics of low frequency oscillation signal we can get advantage and disadvantage of these algorithms.It is important to remember that power system is actually a high-order time-varying nonlinear system. Only under certain circumstances can it be simplified to linear or time-invariant systems. Although ambient condition is reasonably molded as a linear system, for system response following some events, nonlinearities play an important role in the measured data. HHT is a new type of nonlinear and non-stationary signal processing method. Compared with other methods, HHT has absolute advantage of analyzing low frequency oscillation signal because the power system responses following system disturbances contain both linear and nonlinear phenomena.Nevertheless, the traditional methods, whether FFT or WLT, etc. the signals are approximately processed as linear signal when analysis non-linear and non-stationary signals. This feature is the main advantage of HHT algorithm, which is also widely used by the reasons. Secondly, HHT method is adaptive, which means that can be adaptive extracted from the signal decomposed by EMD itself. It is based on an adaptive basis, and the frequency is defined through the Hilbert transform.Consequently, the base of Fourier transform is the trigonometric functions, the base of wavelet transform requires pre-selected. Therefore, HHT has completely adaptability. Third, it is suitable for analysis mutation signal. Due to the Heisenberg uncertainty principle constraint, many traditional algorithms must be satisfied the product of frequency window by time window is constant. This property makes these algorithms cannot achieve high precision both in time domain and frequency domain at the same time. Nevertheless, there is no uncertainty principle limitation on time or frequency resolution from the convolution pairs based on a priori bases. For these reasons, it can be said applying HHT method to dealing with power system oscillation signal is a good choice. However, it is still have some issues need to be resolved carefully.To ensure accurate monitoring of system dynamics with noise-polluted WAMS measurements, serval key signal-processing techniques are implemented to improveHHT method in this research: Data pre-treatment processing, the boundary end effect problem caused by the Empirical mode decomposition(EMD) algorithm and the boundary end effect problem caused by Hilbert transform based on Auto-Regressive and Moving Average Model (ARMA). There are six methods: a). polynomial extension method, b). slope method extension method, c). parallel extension method, d). extreme point symmetric extension method, e). mirror method f). Boundary local characteristic scale extension methods are used to inhibit the boundary end effects, which results in a serious distortion in the EMD sifting process.Furthermore, an integrated scheme for the monitoring and detection of low-frequency oscillations has been developed based on HHT algorithm for oscillation analysis in CampusWAMS projects. By analyzing the real-time synchro-phasors, the proposed scheme is competent to identify the characteristics of the low-frequency oscillations in real-time.Third, this dissertation presents an estimation algorithm method based on enhanced HHT for the parameters of a low frequency oscillation signal in power system.In the end, the developed scheme is tested with simulated signals and measurements from CampusWAMS. An oscillation monitoring system based on real-time wide-area measurements from PMUs is established. It can determine the center rage frequency of the concerned mode automatically and accurately, which is then be used to determine the parameter of the extraction. The extracted mode frequency, damping and mode shape can be detected by this oscillation monitoring system. The results have convincingly demonstrated the validity and practicability of the developed scheme.