Optimization of denitrification carbon source, nitrification aeration and coagulants dosing with artificial intelligence in full-scale municipal wastewater treatment

Conventional AI modeling is often faced with inherent nonlinearity and complex dynamics of industrial wastewater treatment. This work presents a successfully implemented data driven machine-learning AI for process optimization in full-scale wastewater treatment. Tapping on real-time detections from sensors/probes and data transfer through Internet of Things, the proprietary AI realized autonomous fine-tuning of control strategies across anaerobic-anoxic-aerobic (A2O) aerobic tank and final effluent discharge tank for chemical oxygen demand (COD) removal, across A2O anoxic tank and Anoxic Filter for nitrogen removal, and across settler and clarifier for phosphorus removal. Integrating synergistic control and predictions, the AI optimized feed loading, enhanced process efficiency and cost-effectiveness while reducing reliance on human control. Respective removal of COD, NH₃-N, TN, TP and suspended solids up to 97%, 100%, 98%, 100% and 99% rendered discharge compliance. AI optimization reduced over 8%, 11%, 34% and 66% in aeration electricity, sodium acetate carbon source, polyaluminum chloride and polyferric sulfate coagulants. The AI has a complete competitive edge over manual operation in all aspects of process control, resulting in an attractive yearly cost savings of USD206,500, and emissions mitigation of 1,741 tCO₂e/yr from reduction in chemical and electricity consumptions. The implementation revealed that the novel self-adaptive AI can intelligently deal with complex feed dynamics and indefinite dependencies within intricate data. The innovation offers a pointer for sustainable wastewater treatment through maximizing human resources, reducing chemicals, electricity and carbon footprint. © 2025 The Author(s)