Experimental characterization and surrogate modelling of a hollow-fiber membrane liquid desiccant dehumidifier for outdoor air treatment across diverse climates

Humidity control in ventilation air is necessary for indoor air quality. However, membrane assisted liquid desiccant dehumidification for Dedicated Outdoor Air Systems (DOAS) still lacks experimentally based guidance on the selection of surrogate models and on staging requirements in different climates. This study addresses this problem by using an integrated experimental and computational framework for a cross flow hollow fiber polypropylene membrane module operated with lithium chloride solution. A dataset of 181 steady state experiments was collected under conditions relevant for DOAS, with inlet air temperatures of 24.3 to 33.8 °C, humidity ratios of 9.0 to 21.9 g/kg, and solution concentrations of 31 to 35%. The measured moisture removal rate was in the range 0.021 to 0.206 g/s. The sensible effectiveness was 0.191 to 0.605, and the latent effectiveness was 0.134 to 0.416. Nine regression and machine learning methods were compared. Among them, the second order stepwise polynomial regression gave the highest accuracy, with R² values from 0.921 to 0.983 and NRMSE from 3.15 to 5.52%. It also provided explicit equations suitable for hourly simulations. The validated equations were then applied to Typical Meteorological Year data for seven locations representing tropical, subtropical, and continental climates. For the tested module and the accepted operating hours, the climate influenced mainly the required number of stages, while the general applicability remained similar. The new element of this work is not the membrane and LiCl pair itself, but the integrated framework that connects experimental characterization, comparative model benchmarking, and climate specific staging decisions.