Redução do conjunto de casos avaliados em testes de aplicação de relés de proteção utilizando grafo de Gabriel

Abstract

The good performance of protective relay functions is very important to guarantee the correct behavior of the power system. Thus, it is necessary to perform tests in protective relays to ensure the correct operation of protection systems. The IEEE standard C37.233 defines several types of tests for protection systems with different goals. One of them, the so-called application test, is dedicated to evaluate the reliability of a protective relay in a specific situation. It is designed to define the probability of the protective relay to operate properly when it is expected and the probability of the protective relay to not operate properly when it is expected – this procedure must be performed by the final user or under his supervision. The Monte Carlo method is usually performed to do this evaluation and has proved to be an effective procedure for the application test. However, its efficiency has the drawback of requiring a very large number of test cases, which requires more time and labor to perform. As a means to overcome this inconvenience, the present work proposes a new methodology to reduce the amount of tests required for the procedure without changing the process accuracy as the reliability relay estimate. The proposed methodology uses Gabriel graph, a resource available in the area of artificial intelligence for the construction of classifiers, to achieve the objective of the work. A significant reduction in the number of test cases was observed without a significant change in the estimated accuracy of the protective relay settings with the proposed procedure. As a consequence, it is possible to optimize the use of protective relay testing tools such as power system simulators or test cases, as well as Real Time Digital Simulator (RTDS).