Application of machine learning modeling in prediction of solar still performance: A comprehensive survey

Abstract

Being a cheap, simple, and low-energy consumer, solar stills have been introduced by water and energy scientists as an alternative desalination method to fossil fuel-based ones. A wide variety of designs and modifications have been applied to enhance the solar stills' performance, which may be associated with experimental works that require time and cost. Therefore, coupling solar stills with state-of-the-art machine learning is expected to overcome these disadvantages of experimental work. Artificial intelligence models try to build relationships between the input and output data similar to the human brains depending on the given dataset. In light of these, this study carries out a literature review that considers the applications of artificial intelligence in solar stills’ performance prediction. The study covers the most repeated machine learning methods employed for performance prediction, focusing on principles, advantages, limitations, and the mathematical description of each method besides model evaluation criteria. Then, a comprehensive analysis is performed on the solar stills models by classifying them according to the design. The work compares the previous studies within a comprehensive analysis that gives reasons for the authors' findings, highlighting the reasons for the variation between the models' prediction and experimental findings. Accordingly, models with root mean square errors close to zero are highlighted throughout the review.

Topics & Concepts