Projeto de um Controlador Ótimo para Posicionamento de um Painel Fotovoltaico via Aprendizado por Reforço
Authors
Alam Pablo Silva Belfort
Programa de Pós Graduação em Engenharia Elétrica, Universidade Federal do Maranhão, Av. dos Portugueses 1966, São Luís, MA, Brasil.
Yan Ferreira da Silva
Programa de Pós Graduação em Engenharia Elétrica, Universidade Federal do Maranhão, Av. dos Portugueses 1966, São Luís, MA, Brasil.
João Viana da Fonseca Neto
Programa de Pós Graduação em Engenharia Elétrica, Universidade Federal do Maranhão, Av. dos Portugueses 1966, São Luís, MA, Brasil.
ADHDP, Optimal Control, Motor DC, DLQT, Reinforcement Learning
Abstract
With the increase in energy demands in the global electricity sector, the need to use renewable energies such as photovoltaics is clear. The generation of photovoltaic energy depends, among other factors, on the type of photovoltaic panel, the temperature in the module and solar irradiance, therefore to optimize energy generation it is essential to track the position of the module to favor the best generation angle depending on the point maximum power. In this work, an approach for applying reinforcement learning, in the form of Action Dependent Dynamic Heuristic Programming (ADHDP), is proposed for tuning a Discrete controller of the Linear Quadratic Tracker (DLQT) of a motor DC for positioning the panel to obtain the point maximum power, with the lowest possible control effort (energy consumption).
