Comparative Evaluation of Hybrid, Genetic Algorithm, and Particle Swarm Optimization-Based ANFIS Models for Intelligent Thermal Prediction in a Photovoltaic-Assisted Cabinet-Type Solar Dryer

Accurate prediction of drying chamber temperature is essential for improving the energy efficiency and operational stability of solar drying systems. However, the thermal performance of photovoltaic-assisted cabinet-type solar dryers is strongly influenced by variations in solar irradiance, ambient temperature, and relative humidity, resulting in highly nonlinear process behavior. Conventional mathematical models often struggle to represent these complex relationships under changing outdoor conditions. Therefore, intelligent data-driven approaches are required to achieve reliable thermal prediction. This study presents a comparative evaluation of three Adaptive Neuro-Fuzzy Inference System (ANFIS) training approaches for thermal prediction in a photovoltaic-assisted cabinet-type solar dryer. Experimental data obtained under real operating conditions were used to develop predictive models with solar irradiance, ambient temperature, and relative humidity as input variables, while drying chamber temperature was considered as the output variable. Three learning strategies were investigated: conventional Hybrid Learning ANFIS, Genetic Algorithm-optimized ANFIS (GA-ANFIS), and Particle Swarm Optimization-based ANFIS (PSO-ANFIS). Model performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). The results demonstrated that all investigated models successfully captured the nonlinear thermal dynamics of the solar dryer and provided accurate temperature predictions. Nevertheless, optimization-based ANFIS approaches achieved superior performance compared with the conventional Hybrid ANFIS model. Among the evaluated techniques, PSO-ANFIS produced the highest prediction accuracy, exhibiting improved convergence characteristics, better generalization capability, and lower prediction errors. GA-ANFIS also enhanced model performance but remained slightly less effective than PSO-ANFIS. The findings indicate that evolutionary and swarm-intelligence optimization techniques significantly improve ANFIS learning performance for renewable-energy thermal systems. The proposed approach can support intelligent monitoring, predictive control, and energy-management applications in advanced solar drying technologies.