Early warning, intelligent diagnosis, and optimal control of the AAO process under water quality fluctuation shocks: Simulation validation based on a full-scale wastewater treatment plant

This is the authors' abstract. We don't add key points for this paper.

Achieving stable and compliant operation of urban wastewater treatment plants requires effective early warning, process diagnosis, and intelligent control of effluent quality. This study developed a mechanism-data hybrid ML-ASM2d model by integrating Activated Sludge Model No. 2d (ASM2d) with machine learning (ML) for the anaerobic-anoxic-oxic (AAO) process, enabling integrated effluent prediction, interpretable diagnosis, and multi-objective decision-support optimization. The model achieved high-precision same-day prediction of effluent ammonia nitrogen (NH 4 + -N), chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP) (R 2 = 0.90, 0.90, 0.98, and 0.98, respectively), and effectively handled influent fluctuations to provide rapid effluent quality assessment. Shapley Additive exPlanations (SHAP) and Partial Dependence Plot (PDP) analyses identified influent organic load, aeration rate, temperature and pH as core factors, with plant-specific favorable operating domains of 23–25 °C and pH 7.6–7.8. By coupling the prediction model with an improved NSGA-II, a decision-support optimization system was established. For a representative TP exceedance scenario (1.18 mg/L > 1.00 mg/L), the system automatically generated scenario-based strategies: dissolved oxygen in the aerobic zone (DO) increased from 0.60 to 2.64 mg/L, pH from 6.10 to 7.77, internal recirculation ratio (IRR) from 2.49 to 2.18, external recirculation ratio (ERR) increased from 0.64 to 0.92, and mixed liquor suspended solids (MLSS) elevated from 2100 to 4169.52 mg/L. The optimized parameters reduced predicted TP to 0.85 mg/L while maintaining compliance for NH 4 + -N, COD, and TN. This framework establishes an “early warning - intelligent diagnosis, and optimal control” paradigm, providing critical support for compliance-oriented and ultimately resource-efficient wastewater treatment.