Sensitivity analysis of a solar-biomass hybrid cooler with artificial neural network and engineering equation solver

To improve the hybrid renewable cooling system, it is necessary to test the effects of additional key operational factors on its performance. In this study, a comprehensive sensitivity analysis of a solar-biomass thermoelectric-adsorption cooler was conducted to establish the priority of the effects of design and operational parameters on the system efficiency using high-fidelity EES simulation and Artificial Neural Network (ANN) surrogate modelling. The thermodynamic model of the hybrid system was created using the Engineering Equation Solver (EES). A Latin Hypercube Sampling (LHS) plan was employed to create 50 input parameter variations (solar irradiance, fuel biomass, TE current, and adsorption cycle duration). The System Coefficient of Performance (COP) was predicted by the EES model for every combination, producing COP values that ranged from 0.6036 to 0.6307 with a mean of 0.6143. Then, this functionally trained and efficient ANN surrogate model can predict the COP rapidly without hundreds of costly calls to the computationally intensive EES. The ANN model achieved strong prediction performance, with an overall Mean Squared Error (MSE) of 2.12 × 10 − 6 , a correlation coefficient (R) of 0.99478 and a coefficient of determination (R 2 ) of approximately 0.9896 for the complete dataset. Sensitivity analysis performed using ANN examines the effect of the system parameters. Within the investigated design space, permutation-based sensitivity analysis ranked biomass feed rate highest in normalised importance (42.3%), followed by adsorption-cycle time (31.2%), solar irradiance (17.6%) and thermoelectric (TE) current (8.9%). These percentages describe the relative importance assigned by the ANN-based analysis. Overall, this study provides a strong correlation and significant sensitivity analysis. Furthermore, an integrated detailed regression analysis supports the ANN surrogate for an efficient design sensitivity evaluation. Calculations from the study showed that the thermal component (adsorption cycle) operation contributes the most to the overall system efficiency, whereas the impact of the electrical components (TE current) diminishes at high load values. The trained ANN provides predictions on a millisecond timescale, substantially reducing the computational burden associated with repeated EES simulations and supporting rapid design-space screening and sensitivity analysis. The maximum observed COP among the 50 simulated cases is 0.630744608 at run 18, corresponding to a biomass rate of 0.000222222 kg/s and an adsorption-cycle time of 81.99 min. The research findings offer a platform for maximum efficiency optimization for future experimental and smart hybrid cooling system development.