Performance evaluation and machine learning-based prediction of PCM-integrated solar chimney drying for black dates

This study presents a comparative evaluation of black date drying using three solar-based methods: open sun drying, a solar chimney dryer, and a solar chimney dryer integrated with phase change material (PCM). In parallel, machine learning (ML) models were employed to predict and optimize system performance. Experimental findings reveal that the PCM-integrated solar chimney significantly outperformed conventional approaches, achieving peak thermal and drying efficiencies of 49 % and 59 %, respectively, compared to 20 % for open sun drying and 41 % for the standalone solar chimney. The latent heat storage of PCM extended effective drying into late hours, sustaining 25 % efficiency at 16:00 h against only 11 % under open sun drying. Among the tested ML models—multilayer perceptron (MLP), random forest (RF), and support vector regression (SVR)—the MLP demonstrated the highest predictive accuracy (training: RMSE = 0.85, R² = 0.92; testing: RMSE = 1.10, R² = 0.90). Feature importance analysis further identified solar irradiance and airflow as dominant parameters governing drying performance. By integrating PCM-based thermal management with AI-driven prediction, this work establishes a scalable, energy-efficient drying solution to mitigate agricultural post-harvest losses, directly supporting global initiatives on sustainable food processing and renewable energy utilization.