Modeling of Electric Power Transformer Using Complex-Valued Neural Networks

Abstract

Accurate simulation of a power grid requires use of detailed power equipment models in order to reflect maximum of complex dynamics occurs in the grid. Conventional approaches are not always sufficient to fulfill necessity of meticulous description of processes in power devices. Existence of physical difference even between devices of exactly the same type pulls the accuracy of the whole grid simulation using one model for each type of equipment down. A completely new approach of power equipment modeling – modeling based on Complex-Valued Neural Networks (CVNN) – gives an opportunity to build a high-quality models which are able to track dynamics of grid devices. The nature of the approach makes it relatively easy to build models of all electric network devices even individually taking into account the uniqueness of each one. Power transformer, being quite common and, generally, complicated nonlinear element of power grid, has been chosen for demonstration of CVNN method. Results obtained from this work show that application of CVNN in power engineering modeling appears as quite promising method.