As a result of the increase in the number of smart buildings and advances in technology, energy consumption in buildings has become increasingly important. The estimation of energy consumption in buildings is critical for energy efficiency. Accurate estimation of photovoltaic (PV) solar power plant efficiency is crucial for optimizing the performance of renewable energy applications. In this study, advanced machine learning regression models such as XGBoost, CatBoost, LightGBM, AdaBoost and Histogram-Based Gradient Boosting are used to predict PV efficiency based on ten internal features (Open Circuit Voltage (Voc), Short Circuit Current (Isc), Maximum Power (Pmpp), Solar Irradiation Spread (SIS), Maximum Voltage (Vmpp), Maximum Current (Impp), Fill Factor (FF), Parallel Resistance (Rp), Series Resistance (Rs), and Module Temperature (Tm)) of PV module measurements from the Utrecht University Photovoltaic Outdoor Test Facility. As a result, CatBoost outperformed the others, achieving the lowest prediction error MSE of 0.002 and the highest R² value of 0.90. To interpret the model’s predictions, we applied Explainable Artificial Intelligence techniques, in particular SHAP and LIME, which identify key features affecting efficiency and increase model transparency. The integration of these methods provides valuable insights for PV solar power plant design and optimization.