Improved Modeling of Solar Flash Desalination Using Support Vector Regression

Abstract

Accurate prediction of heat-transfer rates in condensers is a challenging task because of phase-change dynamics. This is further complicated if noncondensable gases are present since they tend to form an insulating layer around heat-exchange surfaces. This study examines the utilization of support vector regression in predicting the preheat temperature of seawater exiting a condenser upon its flashing in a vacuum chamber to produce fresh water. Gasses dissolved in seawater are released but not condensed. Thus, system vacuum and heat transfer slowly erode with time due to this accumulation of noncondensable gasses. The preheat temperature is modeled in this study as a function of system vacuum, seawater flow rate through the condenser, and flashed vapor temperature destined for condensation. In comparison with the least-squares polynomial method, the results indicate that support vector regression can predict the preheat temperature much more accurately, resulting in a better performance evaluation of the entire solar desalination system.