Accurate prediction of preheat temperature in solar flash desalination systems using kernel ridge regression

Thermal desalination systems consist of phase-change operations to separate freshwater from bulk seawater. Solar desalination systems involve partial vaporization of seawater, known as flashing, using solar heat then condensation of flashed steam to produce fresh water. The latent heat of the condensing steam is usually utilized to preheat seawater, thus increasing the energy efficiency of the desalination systems. Evaluating the feasibility of a solar desalination system requires accurate determination of seawater preheat temperature exiting the condenser to enter the evaporator. Determining this temperature is very challenging due to the complicated phase-change dynamics and the existence of noncondensable gases in the condenser that were dissolved in seawater. The preheat temperature depends on several factors such as seawater flow rate, system vacuum, and flashed vapor temperature. This study utilizes the kernel ridge regression (KRR) method to predict the preheat temperature as a function of these factors. KRR is an efficient algorithm that utilizes kernel methods to model highly nonlinear relations. The results show the prediction of preheat temperature is highly accurate and it outperforms the state-of the-art regression model of support vector machines in terms of efficiency.