采用改进Prony算法的电力系统故障暂态信号分析

Analysis of power system transient signal based on improved Prony algorithm

为提高对含有谐波、间谐波和衰减直流分量的电力系统故障暂态信号的分析精度,提出基于改进Prony算法的暂态信号分析方法。Prony算法的模型具有能较准确描述故障暂态信号特征、直接提取信号频率的优点。先采用差分算法滤除衰减直流分量并对高频信号进行放大以提高Prony算法的分析精度,利用Prony算法对信号中含有的频率分量进行估计,以确定神经网络的神经元个数和训练的初始值。将各频率分量的频率作为神经网络训练待定的权值,同时估计各频率分量的频率和幅值。仿真结果证明了所提方法的快速性和有效性。

A method of power system transient signal analysis based on improved Prony algorithm is proposed to improve the analysis precision for the transient signal containing harmonic,inter-harmonic and decaying DC components. The model of Prony algorithm can describe the characteristics of transient signal more exactly and acquires the signal frequency directly. The differential algorithm is used to wipe off the decaying DC component and magnify the high frequency components to improve the analysis precision. The frequency components of signal are assumably estimated based on Prony algorithm to determine the amount of nerve cell and the initial values of neural network. Each frequency is treated as the weight to be adjusted and the amplitude and frequency of each frequency is estimated. Simulative results demonstrate the high accuracy and rapid convergence of the proposed algorithm.