基于改进形态滤波与TLS-ESPRIT算法的电力系统低频振荡模态辨识

摘要

针对广域测量低频振荡辨识过程中的噪声干扰和定阶问题，提出一种高精度低频振荡模态辨识方法。该方法基于粒子群优化算法（PSO-GA）设计广义形态滤波器的加权参数，改进后的滤波器可以较好去除噪声；将低频振荡信号通过该滤波器滤波后再使用改进总体最小二乘法-旋转不变技术（TLS-ESPRIT）算法进行模态辨识，可以准确获得各个模态参数。对于辨识算法的定阶问题，把奇异值差值与最大奇异值比值引入到TLS-ESPRIT算法中，采用该方式进行系统定阶，不仅计算量和受主观因素影响小，而且还可以提高辨识效率以及辨识的准确性。通过系统模型仿真以及电网实际案例证明提出的方法能够较快速准确地辨识低频振荡参数，且在抗噪性及辨识精度方面有较大的优势。%A high-accuracy and low-frequency oscillation mode identification method was presented to compensate the weakness of the existing means, especially for the noise jamming and order determination in low-frequency oscillation identification process of wide-area measurement. Weighted parameters of generalized morphological filter were designed via the method based on PSO-GA, which can effectively eliminate the noise. Besides, parameters of each mode can be obtained accurately by filtering low frequency oscillation signal by filters and using the improved TLS-ESPRIT algorithm for identification. For order determination of identification algorithm, not only are the calculated amount and influence of subjective factor small, but also identification efficiency and accuracy can be improved by introducing the specific value of singular value difference and maximum singular value to TLS-ESPRIT algorithm for order determination of system. System model simulation and practical grid case show that the proposed method can quickly and accurately identify low frequency oscillation parameters, and it has greater advantage in noise immunity and identification accuracy.