Inteligência artificial como solução para o problema de restauração em redes inteligentes de distribuição de energia elétrica.

Abstract

Currently, the electricity sector has a direct impact on the country's economic and social development. The disproportionate technological advance contributed to the complexification of the network, imposing a need for change on the electricity sector. Today, the consumer has a low power factor, non-linear loads and can produce and sell energy on the same grid. Elements such as electric vehicles and distributed generations change the load profile of the consumer, which was the basis for the design of the current electrical system. Therefore, they can generate instability, improperly act the protection, inject harmonics, impair the radiality of the system, among other problems. This scenario can leverage the number of faults, threatening power quality. Faults are one of the biggest problems managed by concessionaires, as they are inevitable and most of the time unpredictable. This context, combined with economic interests, drives investment in research in order to analyze or develop techniques that seek to improve administration, automate and increase the resilience of the network. As a solution, the concept of smart grid appears in the literature. One of the main characteristics of an intelligent network, the focus of this work, is the self-healing system, which aims to raise the level of reliability of the network, delivering an immediate corrective response in the event of failures. Therefore, this study aims to present a methodology that, considering the current scenario, can: minimize the collective indicators of continuity, raise the reliability indicators, and optimize the electrical power system through a multi-objective problem, aiming to reduce the damage generated by faults and ensure interoperability. As a solution for the proposed problem, the Particle Swarm Optimization (PSO) meta-heuristic was used, which is part of a branch of artificial intelligence, and can be defined as a population-based stochastic optimization technique. The algorithm was developed in the C++ language, and validated on top of a 54-bar test system from the literature. In structuring the system topology, the concept of graph, adjacency list, together with the dynamic memory allocation function was used. At the end of the work, the results were presented and evaluated.