Analysis Of Prony Based On Improved EMD Filtering For Power System Low-frequency Oscillations

With the national network’s gradually expanding and the high-power excitation system’s widely using, leading the total damping of the power system to reduce or even negative. So it is very easy causing the power system low-frequency oscillations which will do harm to the security and stability of power system. In order to understanding the low-frequency oscillations’mechanism and control method, we need to analyze the power system low-frequency oscillation. In current, the Prony algorithms and empirical mode decomposition (EMD) algorithm is two efficient algorithms of signal analysis in the algorithm for power system low-frequency oscillations.This paper analyzes the principle and characteristics of Prony algorithm in detail firstly: this algorithm can analyze linear signal accurately, but for non-linear signal and the signal with noise, the calculated modal parameters of this algorithm will exist a great error.. Meanwhile this paper analyzes the basic principles and characteristics of EMD algorithm: this algorithm can decompose the non-linear signal or noisy signal into a series of stationary component, then make use of the Hilbert transform will get the each modal parameters accurately, But there is endpoints effect in the decomposition process of EMD algorithms, the component after decomposition will have a fitting errors at both end. Secondly, this paper analyzes the existing achievement of cycle extension, extension based on the slope, parallel extension, mirror extension in improving the fitting error of endpoints, and on the basis of the above analysis, this paper presents an improved EMD continuation algorithm which is based on the waveform correlation coefficient and waveform amplitude relative standard deviation. After comparing the analysis results of examples with other algorithms in Matlab, it is proved to be more effective in restraining the endpoints effect. Thirdly use the analysis method that combining the improved EMD decomposition algorithm with Prony algorithm to get the modal parameters of power system low-frequency oscillation signal. It will not only overcome the problem that there is error in the identification results of Prony algorithm for non-linear signal and noisy signal, but also solve the problem of endpoints effect in EMD decomposition. So it will get the modal parameters of power system oscillation signal more accurately. At last, comparing the results of the new analysis method for linear and non-linear signal, noisy and noise-free signal in Matlab, and using the Simulink platform of Matlab to build a single machine infinite bus system and three9-bus system to simulate the low-frequency oscillations of power systems when single-phase short circuit fault happens, and use the new analysis method to get the modal parameters. It is proved the analysis method that combining the improved EMD decomposition and Prony algorithm to be effective and practicality.