Metodologia para Avaliação de Obras Destinadas à Qualidade do Serviço Utilizando Técnicas de Inteligência Artificial

Abstract

The quality of the electricity distribution service has a great impact on consumer satisfaction and on guaranteeing the concession right for distribution companies. For the concessionaire under study, the main indicators of continuity of service are at levels below the regulatory limits, but due to budget restrictions, the forecast of benefit that the structuring works bring to the continuity indicators must be assertive, for an adequate targeting of investments and decision making. In view of this scenario, a methodology was proposed for the evaluation of works aimed at improving the quality of the service, with the estimate of the benefit associated with the reduction in continuity indicators, using concepts of Artificial Neural Networks (ANN) and Genetic Algorithms (AG). The data used were extracted from the distributor's databases and analyzed to identify the input variables and propose prediction models for the outputs of interest. The historical values of interruptions due to causes were considered as input and the results of continuity indicators associated with the types of works studied form the outputs of the model. The RNA topology used was the Multi Layer Perceptron (MLP). The results obtained by simulating the new methodology showed errors almost 100 times smaller for estimating the benefits of the works compared to the current method used by the distributor.