Estimação fasorial aplicada a relés de proteção numéricos utilizando os métodos de ajuste de curvas e redes neurais artificiais

Abstract

This dissertation proposes two methodologies for the phasor estimation on protective relays in Electrical Power Systems. Firstly, a theoretical introduction of signal processing, the structure of a protective relay and phasor estimation algorithms is presented, including some of the algorithms used on the electrical system, as well as the two proposed methodologies. The first one makes use of the concept of curve-fitting, while the other uses Artificial Neural Networks, both with the goal of performing the real-time estimation of the signal amplitude and phase angle. Secondly, it is made a comparative analysis of the two proposed methods with four well-known and currently used algorithms. This comparison is made through the creation of several test signals with different simulation parameters. From these simulations, six performance indexes are used for the quantitative evaluation of each algorithm, from where the most effective method can be determined through the arithmetic mean of these indexes. Lastly, after all the simulation cases have been presented, a summary of the characteristics of each algorithm is made, based in their numerical results. Then, based on the values obtained on each performance index, the strong and weak points are highlighted, as well as the general classification of each method.