A stochastic optimization framework for condition-based maintenance, crew management, and spare parts logistics in solar energy systems

The rapid deployment of photovoltaic (PV) systems underscored the need for efficient operations and maintenance (O&M) strategies, especially in geographically distributed fleets. This paper presents COMPASS (Condition-based Operations and Maintenance Planning Approach for Solar Systems), a comprehensive stochastic optimization framework that integrates condition-based maintenance scheduling, crew management, and spare parts logistics. The framework leverages predictive analytics on real-world sensor data to generate maintenance signals and explicitly models uncertainties in solar power generation, electricity prices, and maintenance durations through scenario-based planning. A two-stage stochastic mixed-integer optimization model coordinates both hourly and daily decisions to balance preventive and corrective maintenance, crew routing, inventory control, and production planning. To solve the problem, a receding horizon procedure is developed to adaptively incorporate new information over a 30-day horizon. Extensive computational experiments on a real-world PV fleet with 300 assets across 25 locations demonstrate that COMPASS significantly reduces total costs, asset downtime, and inventory costs when compared to the state-of-the-art models.