Optimal cleaning scheduling for large photovoltaic portfolios

Soiling on solar modules stands as a primary source of energy yield loss, causing reflection of radiation. This paper presents a novel cleaning scheduling model for the maintenance strategy of photovoltaic plants, focused on adequately representing the soiling and cleaning processes. We employ a novel methodology based on cleaning sectors to make the model scalable in real-world applications. Computational experiments in a case study of three operating solar power plants in Chile, on three different scales of energy generation, show that the proposed methodology can achieve better results than traditional maintenance models in the literature, increasing revenue an average of 0.8% with respect to an optimized baseline. Additional experiments also show that the proposed model allows for a more efficient use of cleaning resources and makes possible the coordination of multiple large scale power plants, where conventional strategies result on infeasible policies.