ANÃLISE DA ESTABILIDADE TRANSITÓRIA DE SISTEMAS ELÉTRICOS USANDO UMA REDE NEURAL EUCLIDIANA ART&ARTMAP COM TREINAMENTO CONTINUADO

Abstract

This work presents a procedure for analysis oftransitory stability of first oscillation (model classic) ofelectric power systems using a neural network based on thearchitecture ART (Adaptive Resonance Theory), call ofEuclidean ART-ARTMAP with Continuous Training. Thisnetwork is composed for a module Euclidean ART and of amodule ARTMAP, together with a routine of continuoustraining that seeks to turn the most efficient analysis. Thisroutine is constituted an effective possibility of the job ofneural networks in the analysis in real time, destroying theconcept, until then, about the limitations (instability and noplasticity)of the neural networks in the treatment of realproblems that has been checking little application reliability.To illustrate the proposed neural framing, an application ispresented considering an electric system (multimachine)composed of 10 synchronous machines, 45 buses, and 73transmission lines.Index Terms ⎯ Adaptive Resonance Theory, EuclideanART, ARTMAP, Continuous Training, Power Systems,Analysis of transitory Stability.