Can MOF — Isobutane integration enhance adsorption refrigeration cycle? An accelerated approach using active learning and Monte Carlo simulations

Abstract

This study investigates the use of MOF adsorbents with low GWP refrigerant isobutane for a sustainable adsorption-based refrigeration cycle. An innovative active learning-based strategy was used to accelerate the screening process. The combination of a probabilistic surrogate model, trained with a labelled dataset that is iteratively updated by the data query process of an acquisition function, allowed for an efficient exploration of the dataset only in the region of high probability of finding the best MOF rather than the whole dataset. This fusion of active learning with Monte Carlo simulation for labelling the dataset accelerated the screening process by almost 83%. The screening results converged to the highest COP of 0.786 and the highest cooling capacity of 305.9 kJ/kg which is almost 50% higher than the reported value for MOF - isobutane integration. Further, we performed an analysis to find the influence of the largest cavity diameter (LCD) on COP.