AI for smart wastewater treatment plants: A review of physics-informed water quality modeling, optimization, and advanced control.

Wastewater treatment plants (WWTPs) are among the largest energy consumers in cities and are also an important source of global greenhouse gas emissions. The use of artificial intelligence (AI) offers a promising route to make WWTPs low-carbon and smart. This review summarizes research progress over the past decade on AI for effluent quality prediction, process optimization, and advanced control in WWTPs, with an emphasis on how these methods support the development of smart WWTPs. The literature shows that machine learning and deep learning are widely used to predict key effluent indicators. When combined with multi-objective optimization, AI can balance effluent quality, energy use, and carbon emissions, and model-based and reinforcement learning control can further unlock energy saving and emission reduction potential. However, current applications still face limited interpretability, weak transferability across plants, and poor robustness to changing operating conditions. Therefore, this review focuses on physics-informed and other hybrid modeling approaches that embed activated sludge models and process constraints into neural networks, and discusses pathways to couple these models with real-time optimization and control frameworks. Finally, future research needs are identified in data infrastructure, interpretable decision support tools, and full-scale validation studies to enable trustworthy AI-enabled WWTPs and accelerate sustainable and intelligent transformation of the sector.