Large-scale cascade cooling performance evaluation of adsorbent/water working pairs by integrated mathematical modelling and machine learning

Abstract

Efforts to improve the efficiency of adsorption chillers (ACs) catalyzed the invention of cascaded ACs (cACs) with higher overall efficiency that can reutilize the input thermal energy by cascaded stages consisting of a high-temperature stage (HS) and a low-temperature stage (LS). Although the use of water as the refrigerant in cACs is highly attractive, it is an extreme challenging task to identify promising adsorbent/water working pairs from a large number of candidates by a computational or experimental strategy. In this work, the coefficient of performance for cooling (COP_C) and the specific cooling effects (SCE) of over 90 000 cACs based on varying adsorbent/water working pairs were evaluated based on an experimental water adsorption isotherm database (EWAID) by mathematical modelling. Nine cACs with record-breaking COP_C (>1.63) were identified, and it was also revealed that MOFs and zeolites are more potential adsorbents for the HS, while MOFs and COFs are potential candidates for the LS. The relationships between adsorption properties and cooling performance were clearly demonstrated, in which the high water working capacity, I-type water adsorption isotherms with strong adsorbent/water interaction for the HS and V-type adsorption isotherms with the weak interaction for the LS, is favorable for high-performance cACs (i.e., COP_C > 1.5). Besides, the random forest (RF) model of machine learning was successfully executed to accelerate the accurate prediction of the COP_C of thousands of cACs based on adsorbent/water working pairs.