Machine learning model-based optimization of solar-powered direct volumetric steam generation

In recent years solar-powered desalination systems have received much attention as a clean and sustainable solution to freshwater demand. One challenge in the field is to maximize the system’s efficiency. This study focuses on the volumetric direct solar steam generation and provides a model-based optimization using support vector regression and decision tree regression ensemble molding. The model achieves train R2=0.99, validation R2=0.91, and test R2=0.92. For optimization, Nelder-Mead and differential evolution methods are used. Results predict that suspending 0.015 weight-percent of GNP-MWCNT to water achieves the maximum efficiency under 1 kW/m2 radiation.