Energy forecasting and optimization for wastewater treatment facility: A hybrid machine learning – Metaheuristic scenario based approach

Meeting the energy demands of high-consumption facilities, such as wastewater treatment plants (WWTPs), is essential for sustainable urban growth. This study evaluates renewable energy supply strategies through three deterministic models (HOMER software) and a stochastic hybrid machine learning–metaheuristic framework. The deterministic models produced Levelized Costs of Energy (LCOE) of 0.14 $/kW and 0.284 $/kW, while the stochastic model yielded a significantly higher 8.41 $/kW, underscoring the impact of uncertainty on economic feasibility. The hybrid model revealed that battery integration is negligible unless renewable electricity purchase prices rise to 0.45 $/kW, compared to the grid price of 0.036 $/kW (with a hypothetical sellback price of 0.035 $/kW). These findings demonstrate that while deterministic models provide optimistic baselines, the stochastic approach offers a more risk-aware perspective, highlighting the importance of uncertainty modeling in WWTP energy planning. The study contributes a novel methodological framework that integrates machine learning with metaheuristic optimization, offering transferable insights for optimizing renewable integration in high-demand facilities worldwide.