A Reinforcement Learning Model of a Dynamic Solar Panel System for Maximum Energy Harvesting

This paper presents the preliminary results toward developing a reinforcement learning model of a dynamic solar panel system for maximizing solar energy harvesting. We developed a 2DOF intelligent solar panel system actuated by two servomotors: (i) 1DOF was used to adjust the inclination of the solar panel and (ii) another DOF was used to rotate the solar panel with respect to the vertical axis. The system was designed to rotate the solar panel tracking the rotation (position) of the sun. We conducted two separate simulation studies to analyze the static (stress) and dynamic (stability) characteristics of the rotating solar panel system. We conducted an experiment using the solar panel system to harvest energy and compare the harvested energy between two conditions: (i) the solar panel system was set up facing the sun at the beginning and then remained stationary and (ii) the solar panel system rotated at an arbitrary rotational speed tracking the position of the sun. The simulation results proved satisfactory stress distribution over the solar panel system structure and stability of the system. The experimental results showed that the rotating solar panel was able to produce slightly more energy than the stationary solar panel. To improve the system performance, we proposed a reinforcement learning model to learn the rotational speed of the solar panel system following that of the sun that would maximize energy harvesting. The proposed model can be useful to supply solar power to various applications, especially in remote areas such as hydroponic gardening, irrigation, greenhouses, search and rescue operations, military operations, disaster management, etc.