Optimizing dissolved oxygen requirement and energy consumption in wastewater treatment plant aeration tanks using machine learning

Most aeration tanks in wastewater treatment plants (WWTPs) operate at fixed dissolved oxygen (DO) concentrations of 2.0 mg/L. Such control systems do not consider various influent loadings. Thus, it does not respond effectively to fluctuating influent behavior, resulting in energy and cost wastage. This paper proposes a dynamic supervised machine-learning model dependent on real-time influent biochemical oxygen demand predictions and biological treatment oxygen constraints that have not been considered in previous studies for optimizing airblower operation in WWTP aeration tanks. The model is tested against WWTP operating based on the modified Ludzack-Ettinger process. The dataset covers three years and two months of operation, containing 1155 daily readings. The Random Forest and Gradient Boost models predicted the optimum DO concentration by achieving a 1.00 coefficient of determination and 0.01 mean absolute error. The optimization led to a 23 % reduction in energy consumption. Such an intelligence framework can be extended to other air-blower applications and generalized nationwide to optimize energy consumption and cost savings for intelligent WWTPs. The proposed system is distinguished by being automated, intelligent, and following quality assurance and risk management approaches to predict the optimum DO setpoint, suppress the treated effluent quality violation risks, and enhance environmental protection.