Machine Learning Framework for Efficiency and System Optimization in Solar-Powered Absorption Cooling Systems

The growing demand for sustainable cooling technologies has motivated the integration of solar energy into absorption systems. The study focuses on predicting the coefficient of performance (COP) of a solar-powered vapor absorption refrigeration system (VARS) integrated with latent heat energy storage, designed for the climatic conditions of Riyadh, Saudi Arabia. The input parameters considered were incident solar radiation (Ir), evaporator cooling capacity (Qe), absorb efficiency (effa), and heat exchanger load (Qhx), while COP served as the system's output variable. A comprehensive dataset of 8,760 hourly records was analyzed using statistical and machine learning methods. Preprocessing steps included the treatment of missing values, retention of zero values, and dimensionality reduction via Principal Component Analysis (PCA). Correlation analysis revealed strong interdependencies among Ir, Qe, and Qhx, while highlighting absorber efficiency as the most influential factor governing COP. The Random Forest (RF) regression model was trained with 70 % of the data and tested on the remaining 30 %. Results showed outstanding predictive performance, with RMSE values of 0.00032 (training) and 0.00043 (testing), and MAPE values below 0.01 %. The findings confirm the potential of R F as a reliable predictive tool for system optimization, offering accurate forecasting and practical insights for improving absorber and heat exchanger performance.