Estimativa da Magnitude das Tensões em Sistemas Elétricos de Potência via Redes Neurais Artificiais

  • A. Bonini Neto Faculdade de Ciências e Engenharia (FCE), Universidade Estadual Paulista (UNESP), Tupã, SP
  • A. L. Criscimani Faculdade de Ciências e Engenharia (FCE), Universidade Estadual Paulista (UNESP), Tupã, SP
  • W. P. L. dos Santos Faculdade de Ciências e Engenharia (FCE), Universidade Estadual Paulista (UNESP), Tupã, SP
  • J. C. Piazentin Faculdade de Ciências Agronômicas (FCA), Universidade Estadual Paulista (UNESP), Botucatu, SP
  • D. A. Alves Faculdade de Engenharia de Ilha Solteira (FEIS), Universidade Estadual Paulista (UNESP), Ilha Solteira, SP
  • C. R. Minussi Faculdade de Engenharia de Ilha Solteira (FEIS), Universidade Estadual Paulista (UNESP), Ilha Solteira, SP
Keywords: Artificial intelligence, Continuation power flow, P-V curve, Maximum loading point, Mathematical modeling

Abstract

Loading margin evaluation is one of the essential tasks of power system voltage stability analysis. Conventional methods for loading margin calculation are based on continuation power flow techniques. Recently, there is growing interested to apply artificial neural network (ANN) techniques to rapidly predict the loading margin. However, traditional ANN learning algorithms usually suffer from excessive training or tuning burden and unsatisfactory generalization performance. In this work, an ANN to estimate the voltage magnitude is presented, as well as all P-V curves of electrical power systems with reduced training time. From the results, ANN presented good performance, with a mean square error in training below the specified value. Of the samples that were not part of the training, the network managed to estimate 100% of the values of the magnitude of the voltage within the established range, with residues around 10-4 and a percentage of correctness between the desired and obtained output of 99.6%.
Published
2022-11-30
Section
Articles