Integrated Power Generation and Maintenance Optimizationfor Photovoltaic Systems: A Comparative Study ofMachine Learning Approaches

Efficient production planning and maintenance scheduling are critical for the optimal operation of photovoltaic (PV) systems, particularly under uncertain demand and environmental conditions. This paper presents an integrated optimization framework that simultaneously addresses power generation and maintenance planning for a PV system. A comparative analysis is conducted using Support Vector Regression (SVR) and Artificial Neural Networks (ANN) to forecast periodic power generation over a finite planning horizon. The proposed method incorporates system degradation through a reliability-based maintenance model using Weibull distributions. By linking production forecasts with maintenance scheduling, the approach minimizes the total cost of energy production, storage, and system upkeep. A numerical case study based on the Sokoto solar power plant in Nigeria demonstrates the effectiveness and cost-efficiency of the integrated strategy.