The performance of biological wastewater treatment processes directly impacts water resource recycling and ecological safety. This year-long study compared full-scale wastewater treatment plants (WWTPs) using either the anaerobic/anoxic/aerobic (AAO) or modified Bardenpho process. By integrating water quality analysis with 16S rRNA sequencing, we examined how process type, influent quality, and seasonal factors affect microbial communities and treatment performance. Systems with high chemical oxygen demand (COD) and biochemical oxygen demand (BOD)/COD influent exhibited the best pollutant removal performance, with average nitrogen and phosphorus concentrations in the effluent as low as 7.0 mg/L and 0.1 mg/L, respectively. Optimizing a 1:9 influent distribution ratio between the pre-anoxic and first anoxic zones in the modified Bardenpho process increased total nitrogen (TN) removal efficiency by an average of 14 percentage points compared to the AAO process. Additionally, the modified Bardenpho process identified 1100 bacterial genera, indicating a more complex and stable community. Influent water quality had the most significant impact on microbial communities and treatment efficiency, followed by seasonal factors and process type. This study provides theoretical and data support for the optimization of wastewater treatment processes and seasonal regulations.
Deploying artificial intelligence (AI) in drinking water treatment plants (DWTPs) requires frameworks that balance algorithmic deployment with operational safety. To explore an offline evaluation approach prior to direct online control, this study developed a Simulated Closed-Loop Optimization (SCLO) framework. This framework features a bidirectional interaction mechanism for iterative dosage adjustment. Multiple machine learning algorithms were evaluated, with the Generalized Regression Neural Network (GRNN) selected to drive the dual surrogate models, yielding predictions for effluent turbidity (R2 = 0.9347, RMSE = 0.0894 NTU) and coagulant dosage (R2 = 0.9490, RMSE = 1.2660 mg/L). To evaluate mechanistic reliability, SHapley Additive exPlanations (SHAP) analysis was integrated to quantify the algorithmic logic. The analysis indicated that dosage decisions were primarily driven by influent pH and flow rate, which aligns with physical-chemical coagulation principles. Operating within this transparent environment, the SCLO strategy computationally reduced average coagulant consumption by 15.7% in offline simulations while maintaining effluent compliance. These findings suggest that the SCLO framework holds the potential to serve as a decision-support baseline for future DWTP operations.
Life cycle assessments of drinking water treatment plants (DWTPs) typically rely on single-year data or hypothetical scenarios, leaving temporal variability and facility-specific trade-offs poorly understood. This study conducted a cradle-to-gate life cycle assessment of two full-scale DWTPs serving Shanghai over 24 consecutive months, evaluating eight ReCiPe 2016 midpoint impact categories. The plants differ structurally: DWTP-A treats estuarine water using CO₂ pretreatment and polyaluminum chloride, while DWTP-B treats degraded inland reservoir water through heavy alum and ammonium sulfate dosing. Results revealed an asymmetric trade-off between the two facilities. DWTP-B achieved a lower climate change burden (0.396 kg CO₂-eq/m³) than DWTP-A (0.411 kg CO₂-eq/m³) and lower carcinogenic human toxicity, yet DWTP-A outperformed DWTP-B across the remaining six categories, with the marine eutrophication gap reaching nearly fourfold. Wilcoxon tests confirmed all plant-level differences were statistically significant, and six of eight categories remained stable across seasons and years. Grid electricity was the dominant driver of climate impact, while divergent coagulant and ammonium sulfate supply chains drove toxicity and eutrophication. A scenario substituting full photovoltaic power cut carbon emissions at DWTP-A by up to 61% but raised ecotoxicity by up to 66%, reflecting the upstream burdens of solar panel manufacturing and illustrating environmental burden shifting between impact categories. These findings show that environmental optimization in urban water treatment cannot follow a single template, and that utilities need facility-specific strategies, tailored to raw water source and treatment chemistry, to avoid solving one problem while creating another.