Machine learning approaches for predicting energy and exergy efficiency in solar still

Abstract

The global water crisis has made the production of fresh drinking water crucial for human health. Solar desalination is recognized as one of the simplest and most accessible methods for generating drinking water.This research examines how environmental conditions throughout different months of the year affect the energy efficiency and exergy of the Double-slope solar desalination plant. This study further explores the energy efficiency and exergy of solar panels through machine learning algorithms aimed at enhancing their performance. A comprehensive database was created by solving thermodynamic equations, followed by an evaluation of five machine learning models: multilayer perceptron (MLP), MLP BAGGING, EXTRATREES, KNEIGHBORS, and RANDOM FOREST, along with a deep neural network model for regression tasks. The findings indicated that August is identified as the month with the highest water production, peaking between 1:00 and 2:00 PM, which aligns with maximum solar radiation. Changes in wind speed around 8 m/s can increase water production by approximately 14.46 L/s. A ninefold increase in water level can reduce water production by about 3.81 %. Reducing the slope angle from 50 to 10 degrees can increase water production by 1.623 %. Exergy efficiency changes range from 27 % to 30.5 %. It is evident that exergy efficiency is lower than energy efficiency, with the lowest average exergy efficiency recorded in June and the highest observed in August. It was determined that radiation intensity, ambient temperature, tilt angle, and water level are the key factors influencing energy efficiency and exergy, with radiation intensity being the most significant. The findings indicated that the MLP BAGGING model achieved the highest accuracy among the models tested, with R2 = 0.999.