基于小波去噪和随机子空间算法的广域低频振荡估计

Wide-area Low-frequency Oscillation Analysis Based on Wavelet Denoising and Stochastic Subspace Identification

电网规模的日益扩大使得广域低频振荡成为电力系统稳定运行中备受关注的问题之一,提出了一种利用小波软阈值去噪技术,首先对电力系统低频振荡数据进行预处理,然后采用随机子空间算法提取低频振荡信号特征的分析方法。该方法直接利用在线量测数据识别出系统的低频振荡及其特征参数,有效地克服Prony算法、自回归滑动平均算法及希尔伯特-黄等算法受噪声、系统实际阶数的影响大以及单一随机子空间辨识算法难以处理非线性、非平稳振荡信号的缺点。数值仿真及实例分析均验证了基于小波预处理技术的随机子空间算法在电力系统低频振荡分析中的可行性。

With the development of power grid, the inter - area low - frequency oscillations become one of the increasing con- cems in the stable operation of power system. A method using wavelet soft threshold denoising technology is proposed. Firstly, the low - frequency oscillation data of power system is preprocessed, and then the characteristics of low - frequency oscillation signal are extracted by stochastic subspace algorithm. This method identifies the low - frequency oscillation and its characteris- tic parameters based on on -line measurement data directly, and effectively overcomes the defects that Prony algorithm, auto - regressive and moving average （ARMA） algorithm and Hilb - Huang transform algorithm are influenced by noise and actual order number of the system as well as the shortcomings that it is difficult for single stochastic subspace to deal with nonlinear and non - stationary oscillation signals. The feasibility of applying the proposed method to the analysis on low - frequency os- cillation of power system is verified by numerical simulation and instance analysis.