基于形态学分形维数的输电线路故障选相方法

输电线路故障信号是一种典型的非线性信号，分形几何理论为描述非线性故障信号的特性提供了一个有力的分析工具。针对传统分形维数的局限性，本文提出了一种基于局域均值分解（local mean decomposition，LMD）-形态学的分形维数-Elman神经网络的输电线路故障选相新方法。该方法通过对故障电流进行相模转换后，对单一线模分量进行LMD分解得到若干乘积函数（product function，PF）分量，然后选取前4个PF分量进行数学形态学的分形维数估计，最后形成特征向量作为Elman神经网络的输入参数。仿真试验表明：提出的故障分类识别方法能快速、准确地识别各类故障，并且不易受故障初始角、故障位置和过渡电阻的影响，与传统的BP神经网络相比，Elman神经网络具有更好的效果，为准确判断输电线路故障选相提供了一种快速有效的新方法。

Faulty Phase Identification of Transmission Lines Based on Morphology Fractal Dimensions

The fault signal of transmission lines is typically nonlinear, and the fractal geometrical theory provides an efficient tool to describe its features. In order to overcome the deficiency of the traditional calculation method of the fractal dimensions, this paper presents a new method for faulty phase identification of transmission lines based on local mean decomposition （LMD）-morphology fractal dimensions-Elman neural network. Firstly, it decomposes single line mode component into some product function （PF） by LMD after the phase-to-analog conversion of the fault current;Then it calculates the morphology fractal dimensions value by the first four PFs;Finally, it inputs the value into the Elman neural network as a new eigenvector to characterize the fault type. Simulation results show that this method can effectively extract the fault types and is also insensitive to different fault initial angles, dis-tances and resistances. Moreover, compared with BP network, Elman neural network has better results. Thus the proposed method was verified as a fast and effective technique for the accurate identification of faulty phase of transmission lines.