Forecasting Weather and Energy Demand for Optimization of Renewable Energy and Energy Storage Systems for Water Desalination

The increasing intensity and frequency of water scarcity, carbon emissions, and climate risks pose critical challenges necessitating increased uptake of and a paradigm shift to energy- and climate-smart water desalination processes. This study employs metrics and a decision framework to enable and accelerate the energy efficiency, decarbonization, and cost-effectiveness of water desalination processes. As an essential step, we analyze various Renewable Energy (RE) sources, such as photovoltaic, wind, concentrated solar power, geothermal, and hydro energy; in addition, we examine battery storage systems to address the intermittency challenges associated with solar and wind energy. The feasibility of these diverse RE systems was assessed at four (4) mid-to-large scale U.S. desalination plants using operating plant and weather/environmental data, establishing optimization functions and constraints. In this research, to facilitate a comprehensive Energy Management System (EMS), we align RE generation with the anticipated energy demand of the plants. Machine Learning (ML) models, including SARIMA, Random Forest, XGBoost, and Gradient Boosting, are employed for forecasting water production, energy consumption, and long-term weather. The results show that Artificial Intelligence (AI) models, notably Gradient Boosting and an innovative XGBoost average method, demonstrated high accuracy in forecasting critical variables for RE systems in water desalination, with a normalized Root Mean Square Error of less than 10% for key metrics. This study can serve as a basis to optimize the mix of hybrid RE systems to minimize cost and carbon emissions.