Restabelecimento multiobjetivo de sistemas de distribuição através de método híbrido de otimização

Abstract

This dissertation proposes a hybrid multiobjective optimization methodology for the restoration of electric energy distribution systems. To treat the mixed and combinatorial nature of the problem, due to the presence of discrete variables, associated to decision-making for reestablishment, together with continuous variables of the electric network state, optimization technique meta-heuristics bioinspired for binary Particle Swarm is proposed. The binary nature of the optimization algorithm proposed does not require the adoption of rounding strategies for discrete variables, which may compromise the effectiveness of the method. A genetic algorithm is proposed for the case of a candidate solution violate the voltage limit in any electrical network bar, where it is possible to cut the load to meet the voltage limits. Multiple objectives are considered for the restoration problem, including minimizing the energy not supplied during this process, attending to priority consumer, minimizing technical losses and minimizing the number of necessary switching operations. In order to address these different objectives, which may be even conflicting, such as the minimization of energy not supplied and the minimization of the number of maneuvers, the Pareto dominance method is used embedded in the particle swarm algorithm, generating the algorithm hybrid. In addition, principles of fuzzy logic are incorporated to define the Pareto fronts and obtain candidate solutions that are competing and alternative to the problem. The advantage of using the Pareto method with fuzzy logic is the possibility of obtaining competing candidate solutions, instead of a single solution, in order to support decision making by the system planner and / or operator, according to the reality and priorities of each distributor. The optimization model for restoration includes important constraints to the problem, such as the radiality of the electric distribution network and limits of nodal voltage and currents in the stretches of the network. Three systems are used to validate the proposed methodology, including a real network of the Southeast region of Brazil.