Aplicação de redes neurais na proteção de sistemas de distribuição ativos

Abstract

Conventional power distribution systems are radials in nature, characterized by the unidirectional flow of energy, starting from the substation to the final consumers. The increase of distributed generation causes several impacts in the distribution systems and it represents a new challenge in the way of operating and protecting these networks. One of the significant impacts occurs on overcurrent protection in distribution feeders. The distributed generator can cause delay in the protection or even desensitizing the relays. Its effect on protection will depend on size, type and place of installation. This work presents the development of a protection scheme for active distribution systems based on artificial neural networks. This scheme suggests the use of three stages of feedforward artificial neural networks to assist in the protection of active distribution feeders. The first one is able to detect the fault on the system. The second indicates in which feeder the fault occurred and the last one is able to precisely locate the fault along the faulty line. Results show that the proposed scheme is satisfactory as a solution to the identified problems.