Modelo de programação não-linear para a reconfiguração e alocação de geração distribuída em redes de distribuição de energia elétrica

Abstract

The Reconfiguration of Electric Power Distribution Systems (REPDS) is a traditional problem and of high mathematical complexity, since the operating state of the disconnecting switches enables the reduction of total operating costs, as well as the reduction of technical losses, so that the minimum voltage limits necessary to meet the loads are respected. The REPDS problem seeks to find the optimal combination of the states of the existing intercon-nection switches, so that, simultaneously, it meets the operational constraints, minimizes the objective function in focus and maintains the radiality, whereas the study of the allocation of the Distributed Generation (DG) it seeks the best connection location for the generators dis-tributed in the system in such a way that it can also minimize operating costs and total active power losses, complying with the voltage levels established by current national regulations. In this way, this work presents a methodology to solve the REPDS problem in the presence of DG through a Mixed Integer Nonlinear Programming (MINLP) model,developed in the AMPL - A Mathematical Programming Language - programming environment and solved by the commercial solver knitro. Several tests were performed on electrical systems commonly found in the technical literature, to verify the convergence of the proposed model to solve the REPDS problem together with the DG allocation. To guarantee success in the implementa-tions, initially, both problems were solved independently and after finding consistent results, both problems were solved simultaneously. From the results obtained, it was found that the adopted unified model (REPDS with DG allocation) has high reliability and presents optimal solutions that significantly reduce active power losses when compared to the system in its original configuration. However, despite its reliability, the solution of the nonlinear model requires a high computational cost when dealing with a system of 69 buses or higher.