Machine learning and application for modeling and prediction of desalination cost globally

Abstract

This study presents a machine learning model and web-based application to estimate the Optimum Cost-Effective Solution (OCES) for desalination plant construction, defined as the minimum capital expenditure (CAPEX) required for a specific desalination technology at a given location. Using a dataset of over 21,000 desalination projects, refined to 10,314 after data cleaning, key variables analyzed include plant size, technology type, location, and procurement models. CAPEX values ranged from $8000 to $2.56 billion, influenced by plant scale, technology, geographic constraints, and logistical challenges. The study examines reverse osmosis (RO), multi-stage flash (MSF), and multi-effect distillation (MED), identifying RO as the most cost-effective due to its energy efficiency. Traditional cost estimation methods, such as parametric models and rule-based approaches, rely on historical averages, fixed cost coefficients, and expert judgment, often failing to capture complex, nonlinear interactions. Machine learning significantly improves estimation accuracy by identifying hidden patterns in large datasets. Evaluated models, including linear regression, decision trees, and ensemble methods, showed varying accuracies, with the CatBoostRegressor reducing prediction errors by over 50 %. It performed best for small and medium-sized plants, while large-scale plants exhibited greater cost variability. Key findings indicate that larger plants, specialized technologies like thermal desalination, and remote locations increase costs. Procurement models impact CAPEX, with privately funded and public-private partnership (PPP) projects proving more cost-effective than government-only models due to competitive supplier dynamics. Plants using equipment from leading membrane and energy recovery device (ERD) suppliers achieved better cost-to-efficiency ratios. A geographic analysis revealed that the Middle East and North Africa (MENA) had higher plant densities and higher CAPEX due to extreme freshwater demands. This study offers a valuable tool for optimizing desalination project costs to address global water scarcity challenges.