Wastewater treatment plant operation optimization towards energy efficiency and reduced effluent loading based on optimized data-driven models

Traditional wastewater treatment plant operations often face challenges in balancing treatment performance with energy consumption, particularly under dynamic influent conditions. This study presents a novel approach to model wastewater treatment plants by leveraging statistical, data-driven and optimization methods to enhance both classical modelling and recent data-driven modelling approaches. Historical operational influent and effluent data from a full-scale plant was used to accurately predict effluent key performance indicators, such as total nitrogen and orthophosphate concentrations. The optimized data-driven model development methodology resulted in the most accurate models for all output features with model performance indicators of up to R 2 = 0.99 and mean absolute percentage error