Screening adsorbent-working solution pairs for adsorption-driven osmotic heat engines based on experimental water adsorption isotherm database and machine learning

Abstract

Osmotic heat engines have attracted increasing attention in harvesting ultra-low temperature waste heat. In order to fill the gap in the high-throughput computational screening of adsorbent-aqueous salt solution working pairs for adsorption-driven osmotic heat engines, an experimental water adsorption isotherm database is constructed and eight common salt-water solutions are selected to identify the high-performance work pairs with system energy efficiency as evaluation indicator. The relationship between adsorbent properties, adsorbent structure characteristics and system performance is systematically analyzed. Results revealed that high working capacity and moderate adsorption enthalpy of adsorbents and large osmotic coefficients of salts are beneficial to energy efficiency. Adsorbents with larger accessible surface area, moderate available pore volume and critical pore diameter are favorable. Furthermore, regression machine learning is employed for achieving fast and accurate prediction of the system energy efficiency to accelerate screening. Genetic algorithm is adopted to search for the best-performing working pair properties.