Toward Improving Solar Panel Efficiency using Reinforcement Learning

Solar energy offers a pollution free and sustainable means of harvesting energy directly from the sun. Considerable effort has been directed toward maximizing the efficiency of end-to-end solar systems, including the design of photovoltaic cells [15, 26], engineering new photovoltaic architectures and materials [24], and solar tracking systems [4]. Solar tracking is especially important for maximizing performance of solar panels [8, 38, 21]. Given the proper sensors and hardware, a tracking algorithm can compute the relative location of the sun in the sky throughout the day, and a controller can orient the panel to point at the sun, illustrated in Figure 1. Its goal is to minimize the angle of incidence between incoming solar radiant energy and the grid of photovoltaic cells, as in Eke and Senturk [8], Benghanem [3], King et al. [21] and Kalogirou [17].