Hybrid Modeling of Solar Drying of Peppers: Integration of Classical Thermodynamics and a Predictive Random Forest Model

Abstract

Abstract

This study investigated the drying behaviour of green, red and yellow peppers in an indirect forced-convection solar dryer by combining conventional drying-kinetics analysis with Random Forest prediction. Solar irradiance, dryer-air temperature, product temperature and changes in product moisture content were monitored during drying. Irradiance reached 872 W/m², while dryer-air and product temperatures reached 54.6 °C and 56.6 °C, respectively. The air temperature exceeded ambient temperature by more than 21 °C during the reported period. All varieties showed a decrease in moisture content with time, with a higher initial evaporation rate and faster dehydration observed for green peppers. Red and yellow peppers showed broadly similar moisture-content profiles. Six semi-empirical models were fitted to the experimental moisture-ratio data. The two-term model provided the best fit for green peppers (R² = 0.9947), whereas the Midilli model performed best for red and yellow peppers (R² = 0.9908 and 0.9903, respectively). Effective diffusivity, estimated using a simplified solution of Fick's second law, was highest for green peppers (4.943 × 10⁻¹⁰ m²/s), followed by yellow (3.125 × 10⁻¹⁰ m²/s) and red peppers (2.768 × 10⁻¹⁰ m²/s). Random Forest feature importance identified elapsed time (approximately 35%) and product temperature (approximately 28.7%) as the leading predictors of residual moisture ratio. The findings describe varietal differences in solar-drying kinetics and show the potential value of combining semi-empirical drying models with data-driven prediction.