Modeling and prediction of air conditioning condensate generation: Experimental validation and environmental benefits

Air conditioning condensate is a valuable source of water and chilled energy. This study presents an analytical model, experimental validation, and machine learning-based prediction of condensate generation. The models were developed using a dataset of 17,500 data points from air conditioning systems with cooling capacities of 25 TR (tons of refrigeration) and 50 TR, under varying latent heat loads (25% and 35%) and fresh air ventilation rates (10% and 25%). The input parameters included outdoor weather conditions, cooling capacity, latent heat load, and fresh air ventilation. The analytical model was validated against a 33 TR, showing a relative error below 10% and a coefficient of determination of 0.95. The dataset was further used to predict condensate generation with multiple machine learning models, including artificial neural networks, support vector regression, gradient boosting, extreme gradient boosting, and random forest. Among these, the random forest model demonstrated the highest predictive accuracy, followed by extreme gradient boosting, with ensemble-based approaches outperforming other models. Shapley Additive Explanations (SHAP) analysis identified cooling capacity as the dominant factor influencing condensate prediction. Experimental validation of the random forest model confirmed its accuracy, with an error margin below 6 % and a coefficient of determination of 0.98. The life cycle impact assessment revealed that condensate recovery from 10–50 TR AC systems can save 11.77–68.37 kWh/day of electricity in coal-based power generation and 4.2–13.7 kWh/day through chilled water equivalence, demonstrating significant potential for environmental impact reduction. • Air-conditioning (AC) condensate is a potential resource for water and energy. • Condensate generation modeled using analytical and data-driven approaches. • Heat load increase from 25% to 35% raises condensate generation by 30–33%. • Random forest model showed the highest prediction accuracy among other models. • Life cycle assessment shows AC condensate recovery reduces environmental impacts.