Reconhecimento de padrões relativos aos problemas que causam variações de tensão em regime permanente utilizando redes neurais artificiais

Abstract

Due to constant technological advance in the last decades, awareness and information dissemination to the population, globalization and competition of productive sectors and the regulation of the Brazilian electricity sector, the electrical energy quality has gained great importance in the overall scenario, mainly due to the financial impact that the lack of electrical energy quality can bring to the varied classes of consumers. Thus, this work proposes a tool to assist dealers and licensees in the electrical energy distribution regarding the preliminary verification for problems related to long-term voltage variations, without the need of a thorough evaluation of the system, which means that the problem will be simultaneously classified accordingly to the voltage measurement performed on the unit consumer. The tool proposed in this work is based on the utilization of Artificial Neural Networks with feed-forward architecture of multiple layers, known as Multi-Layer Perceptron (MLP) for pattern recognition of problems that cause long term voltage variations in energy distribution systems. This research proposes a simple RNA topology and easy implementation in hardware. Data presented for the network in the training and testing phase were provided by the electricity distribution utility, CPFL (Companhia Paulista de Forca e Luz), which were collected from actual measurements made in consumers units. The proposed tool seeks to identify the cause of problems as soon as there is a consumer complaining, in a quick and efficient way.