基于改进Prony算法的电力系统低频振荡模式识别

提出了一种新的改进Prony算法，该算法将待求振荡幅值作为权值，基于神经网络进行训练，实现对电力系统低频振荡模式的识别。该算法避免了Prony算法在实际计算中矩阵呈病态以及通过矩阵求逆计算幅值和相位时精度不高的问题，克服了传统Prony算法抗干扰较差的问题。仿真结果表明，该改进Prony算法能有效去除干扰，能可靠、准确地识别主导模式，计算量少，适用于识别含有噪声且采样点数多的振荡信号。

Identification of Power System Low Frequency Oscillation Mode Based on Improved Prony Algorithm

A new improved Prony algorithm is presented in which the oscillation amplitude to be solved is served as weight and the trained by neural network to implement the identification of power system low frequency oscillation mode. The proposed algorithm avoids the defects while Prony algorithm is applied in actual calculation, such as ill-conditioned expression of matrix and low accuracy of amplitude and phase calculated by matrix; and overcomes the shortcoming in weak anti-interference ability of traditional Prony algorithm. Simulation results show that the improved Prony algorithm can eliminate interference effectively and identify dominant mode reliably and accurately, besides its calculation burden is light, so the proposed algorithm is suitable to identify the oscillation signals containing noises under multi sampling number.