Specific energy consumption of seawater reverse osmosis desalination plants using machine learning

Abstract

Water scarcity is intensified by population growth, industrial expansion, and limited water resources. Seawater reverse osmosis (SWRO) desalination offers a solution by supplying freshwater, yet its high specific energy consumption (SEC) restricts more comprehensive application. It is crucial to understand the impact of design variables on SEC for improving SWRO efficiency. Machine learning (ML) enables the analysis of complex datasets and predictive modeling, revealing how these variables affect SEC. This study adopts ML to evaluate the SEC of SWRO plants, incorporating recent advancements and offering in-depth analysis to drive improvements in energy efficiency. First, linear regression analysis revealed difficulty in isolating the effect of individual variables due to multiple influencing factors on SEC. Various ML models were tested for predictive accuracy, with the extreme gradient boosting model exhibiting the highest performance. Shapley additive explanations and permutation feature importance analyses identified energy recovery device type and commissioning year as critical influences on SEC, highlighting areas for targeted efficiency improvements. Further analysis with the developed ML model demonstrated the SEC of extra-large SWRO desalination plants with state-of-the-art technologies to be 2.8 kWh/m³. These results highlight the role of ML in evaluating factors influencing SEC and guiding the development of sustainable desalination technologies.