Estudo, análise e implementação de uma metodologia para otimização de rendimento em sistemas compostos por conversores em paralelo

Abstract

This PhD Dissertation proposes a new methodology for efficiency maximization in systems comprising parallel converters. For a given number of system converters, their maximum power and their efficiency curves, the proposed methodology identifies the number of necessary converters and an optimum power distribution so that the maximum efficiency is achieved for all power levels that the system is able to handle. It is shown that the efficiency maximization in systems formed by parallel converters is a nonlinear optimization problem with restrictions and could present multiple local minima or even multiple global minima. To overcome this problem, the proposed methodology employ stages of global optimization, by means of a genetic algorithm, and local optimization, carried out by a a sequential quadratic programming algorithm. The importance of efficiency maximization becomes clear when the systems are fed by renewable sources as wind or photovoltaic, mainly during low-power operations. In this sense, an analysis of the capacity of photovoltaic energy generation is presented for some Brazilian cities, which is built from data bases of meteorological stations installed in the cities of São Martinho da Serra (RS), Ourinhos (SP), Brasília (DF) e Petrolina (PE). From this analysis a method for obtaining the equations of weighted average efficiency is presented for each region according to the procedures established by the IEC 61683 standard. A supervisory control strategy is also proposed to implement the optimal power distributions that results from the application of the proposed methodology. This control strategy is based on the active current sharing between the converters from one or more decision variables. An extensive experimental verification evaluates the dynamic behavior and the employment of the supervisory control strategy to different operating points, with variations both for handled power and for input voltage. For all the analyzed cases, one demonstrates that the proposed methodology is capable to precisely find the power distribution that provides the maximum efficiency for the system. Comparative analysis against a conventional power sharing strategy shows that the proposed methodology allows a significant efficiency improvement of the system, mainly for low-power operation.