Abstract of Thesis presented at COPPE/UFRJ as a partial fulfillment of the requirements for the degree of Doctor of Science (D.Sc.)

Proposal of a Method for Power Transformers Equivalent Circuit Parameters Identification

Gisella Margarita Vizhñay Zambrano

April/2009

Advisors:

Antonio Carlos Ferreira Luiz Pereira Calôba

Department:

Eletrical Engineering

      This thesis presents two methodologies to identify the parameters of an equivalent circuit which represents the electric performance of the power transformer in a wide frequency range.
      The first identification algorithm, based on the Specialized Artificial Neural Networks (SANN), approximates the impedance curve obtained by the Terminal Impedance Test applied on a certain power transformer. The parameters are determined from the estimated impedance curve using Circuit Synthesis Procedure. The SANN was combined with the Genetic Algorithm in order to optimize the initial values of the weights in the first layers for speeding up the convergence of the algorithm.
      Another estimation technique, based on the optimization procedure of the Descendent Gradient (DG), is considered for defining the optimal values of the equivalent circuit parameters of the power transformer, directly from the impedance curve. In this approach, three optimization cases are implemented: a) with one objective function, b) with two objective functions and c) with four objective functions. Each one of those functions relates to the mean quadratic error of the impedance curve and the mean quadratic error of the voltage ratio curve, obtained from the Frequency Response Test.
      The methodologies are validated against impedance and voltage ratio curves obtained from the Terminal Impedance and Frequency Response of a real transformer.