Algoritmos Genéticos Para Obtenção de Modelo de Parâmetros para Módulos Fotovoltaicos de 1, 2 e 3 Diodos

Abstract

Among computational techniques, genetic algorithms have been widely used in system optimization and problem modeling. This work aims to use genetic algorithms to obtain the parameters of models of 1, 2 and 3 diodes for photovoltaic modules. This research is motivated by the difficulty in obtaining, through numerical methods, the parameters of models of photovoltaic modules with 2 and 3 diodes. This model has two exponentials in its equation, which hinders the convergence of the iterative method, due to its expression with rigid characteristics with a numerical solution step that meets the double exponentials. In previous work, the use of trust region methods and genetic algorithms was successfully proposed as an initial condition for the Newton Raphson method to obtain model parameters with 1 diode of photovoltaic modules. In this research, the models with 2 and 3 diodes did not converge to the Newton Raphson method due to the special characteristics of the equations in the models. In order to simplify this task, the proposal is to use genetic algorithms to obtain these parameters, without the need to use more complex specific numerical methods cited in the literature that are suitable for convergence. Although the genetic algorithms do not converge to the same results for each solution of the same problem, it was observed that the difference is very small and does not compromise obtaining the parameters for different solutions. One of the main ways to improve the efficiency of photovoltaic cells is to adjust their operating parameters, such as voltage, current and electrical resistance. This can be done through optimization techniques, which seek to find the ideal values of these parameters for each type of cell. However, the search for these values can be a time-consuming and complex process, as it involves the evaluation of multiple parameters and their interaction with the properties of the materials used. Using data with curves obtained through direct measurements performed on a photovoltaic module and comparing with data via Genetic Algorithms (GA) to obtain parameters for 1, 2 and 3 diodes, it was concluded that the parameters via GA can be used, as they were very close to the real curves. The dissertation describes the equations for modules with 1, 2 and 3 diodes and the parameters obtained through GA and direct measurement results.