FFT结合神经网络的低频振荡主导模式识别

Power system low frequency oscillation dominant pattern recognition based on FFT and neural network algorithm

为提高电力系统低频振荡主导模式识别的抗噪性,提出一种FFT结合神经网络的识别方法.首先,基于加窗插值FFT算法求解各振荡模式的频率及其能量权重;然后利用神经网络分段逼近低频振荡信号,根据相邻两段的幅值变化求解衰减因子;最后拟合求出低频振荡信号的幅值和相位.仿真结果表明,该方法能可靠、准确地识别低频振荡主导模式,与Prony算法相比具有较好的抗噪性.

In order to improve noise immunity of power system low frequency oscillation dominant pattern recognition, a new algorithm based on FFT and neural network is proposed in this paper. Firstly, the frequency and energy weight of the dominant mode can be calculated by using the windowed and interpolated FFT algorithm. Secondly, the neural network method is used to ana- lyze low-frequency oscillation signal, and the damping factor can be calculated based on the varia- tion of amplitude. Finally, the amplitude and phase of low-frequency oscillation can be figured out. Simulation results show that this algorithm can identify the dominant mode reliably and ac- curately, which has better noise immunity compared with the Prony algorithm.