Using Evolutionary Algorithms to Determine Frequency-Dependent Network Equivalents

Abstract

The present paper presents an alternative methodology to determine frequency-dependent network equivalents from the frequency response of electric power networks. According to this methodology, one can substitute part of the power network through a simplified electric circuit, reducing the number of elements to be modeled without introducing any inaccuracy to the precise evaluation of large power networks. The main challenge relays on determining a rational function capable of producing the same frequency response characteristic of the original circuit that would be replaced by the equivalent. The rational function determines the impedance to be used in the network equivalent. The methodology is based on Evolutionary Algorithms (EAs). Basically, different set of values for the rational function parameters are evaluated. During the convergence process, the EAs modify the values for the function’s parameters suitably, reducing the RMS error between the original frequency response and the one produced by the equivalent circuit. The results obtained through this methodology were compared with those obtained through the “Vector Fitting” methodology, which is the methodology commonly used to solve similar problems.