基于RBF和TFT的自适应宽频振荡监测方法

A method for broadband oscillation monitoring based on RBF and TFT

近年来,随着可再生能源的大力发展,电力电子设备的渗透率越来越高,电力系统的振荡特性逐渐呈现出宽频化特征,而现有的广域监测系统(WAMS)面向工频分量,难以满足宽频振荡监测、控制、保护的需求。为此,本文提出了一种基于径向基(RBF)神经网络和泰勒傅里叶变换(TFT)的宽频振荡监测方法,实现了宽频振荡信号的精确估计。首先,利用离散傅里叶变换(DFT)进行初步估计,然后采用TFT精确计算宽频振荡信号的参数。为降低TFT算法的计算量,本文将RBF神经网络用于噪声强度估计,根据噪声大小自适应确定数据窗长。最后,对大量仿真数据及河北沽源和新疆哈密的实测振荡数据进行了验证,结果表明即使在噪声较大时,RBF神经网络的拟合效果也十分出色,文中方法的精确性仍然较高,因此,有望在未来应用于工程实际中。

With the development of renewable resources in recent years,the penetration rate of electronic equipment is getting higher and the oscillation is taking on characteristic of broadband.However,the existing wide-area measurement system(WAMS)mainly focuses on fundamental component,which is difficult to satisfy the demands of monitoring,control,protection of broadband oscillation.In this paper,a method based on radical basis function(RBF)neural network and Taylor-Fourier transform(TFT)of broadband oscillation monitoring was proposed,which realized the accurate estimation of broadband oscillation.Firstly,the discrete Fourier transform(DFT)was used to get a preliminary estimation,and then TFT was used to get accurate estimation of signal.In order to reduce the computation of TFT method,RBF neural network was used for noise intensity estimation,which determined the window length of data adaptively.Finally,the proposed method was verified through a large number of simulation data and filed data of Guyuan,Hebei and Hami,Xinjiang.And experiment results manifest that RBF neural network has extraordinary de-noising effect and the proposed method shows high accuracy,even under low signal-noise ratio(SNR).Hence,it is expected to be applied in real engineering.