基于迭代加权改进BP神经网络的电容式电压互感器变比幅频特性拟合方法

针对电容式电压互感器(CVT)的变比幅频特性曲线拟合误差较大的问题,文中以BP神经网络为基础,提出了迭代加权改进的BP神经网络方法,对CVT的变比幅频特性曲线进行拟合,该方法不仅可以保证拟合曲线的均方误差满足收敛条件,还可以有效降低拟合曲线的最大误差。实例计算表明,该方法使得CVT变比幅频特性拟合曲线的最大误差降低24.45%。

A fitting method of the magnitude-frequency characteristic of capacitance voltage transformer transformation ratio based on iterative weighted improved BP neural network

Aiming at the problem of large fitting error of the transformation ratio magnitude-frequency characteristic curve of capacitive voltage transformer (CVT), this paper proposes an iterative weighted improved BP neural network method based on BP neural network. When the characteristic curve is fitted, this method can not only ensure that the mean square error of the fitted curve meets the convergence condition, but also can effectively reduce the maximum error of the fitted curve. Example calculations show that this method reduces the maximum error of the CVT transformation ratio magnitude-frequency characteristic fitting curve by 24.45%.