npj Clean Water最新文献

Pilot testing using feed water sourced from the Yuma Desalting Plant (~2 g/L) (Arizona, USA), an inland brackish water desalination facility, was conducted using either tight Dupont Filmtec Nanofiltration (NF) NF90 membranes or looser NF270 membranes as integrated, pre-treatment, or brine recovery for Reverse Osmosis (RO). The hybrid configurations that include both NF270 and RO membranes exhibited the highest RO water flux, 37– 41 Lm⁻² h⁻¹, with over 99% salt rejection. However, the cost was strongly influenced by the volume of brine produced compared to the energy consumption, resulting in the lowest cost in the NF270 brine recovery configuration. Both the pilot study and modeling data indicate that NF270 and RO membrane hybrid configurations are an economically viable treatment for water purification in inland areas where brackish water is a prevalent water source.

Large language models (LLMs) have shown significant promise for water and wastewater management. However, current foundation models are not yet reliable. This Perspective outlines a pathway for customizing foundation models into WaterGPTs (specialized LLMs for water and wastewater management). We present key methodologies for adapting foundation models into WaterGPTs, including prompt engineering, knowledge and tool augmentation, and fine-tuning, and they are illustrated through representative examples. Then, we highlight the importance of diverse and ethically sourced datasets to customize foundation models, and we propose strategies for efficiently extracting high-quality information to customize foundation models. Further, we advocate for the development of a secure, informative, and dynamic evaluation benchmark that will guide the creation of more reliable WaterGPT. To illustrate practical LLM deployment in water sectors, we envision a specialized WaterGPT in wastewater treatment plants, which could integrate specific biological/chemical knowledge and advanced tools to manage intricate processes of activated sludge. Collectively, we aim to lower barriers for non-AI water-domain-specific experts and bridge the gap between experimental and computational research in water and wastewater management.

Perikinetic and orthokinetic flocculation are the first steps in drinking water treatment plant (DWTP) and affect all subsequent processes. Leveraging multi-stage water quality parameters, we developed a machine learning (ML) framework for coagulation control that incorporates knowledge embedding (KE) through hyper-parametric constraints on threshold water quality, energy consumption, and economic costs. Random forest (RF) has the best performance among the eight methods with a percentage error of 2.53% and a coefficient of determination of 0.9922. The results of the interpretability analysis show that the model can accurately identify the coagulation demand and balance the removal effect with the energy consumption and economic cost. Through real experimental validation and simulation extrapolation, the RF-KE model can reduce turbidity by 16.36% and dosing cost by 9.64%. This framework reduces economic costs while optimizing water quality through KE and interpretability analyses, providing evidence for the safe and reliable application of future models.

Traditional calibration of Activated Sludge Models (ASM) is often manual, expert-dependent, and inefficient. This study introduces a hyperparameter optimisation framework using Optuna to automate the calibration of the ASM2d model. Built on Python, the model integrates the Tree-structured Parzen Estimator (TPE) for single-objective and NSGA-II for multi-objective optimisation. A 50-day dataset from a full-scale wastewater treatment plant in Shenzhen, China, validates the approach. Compared to traditional methods, TPE reduced average relative errors for TN and COD from 4.587 and 24.846% to 0.798 and 15.291%, respectively, while decreasing iterations by 15–20%. NSGA-II lowered TN and COD errors to 4.72 and 15.17%, further improving to 0.095% and 8.43% with full-parameter tuning. Calibration efficiency increased by 65–75%. By effectively exploring parameter interdependencies, TPE and NSGA-II enhance calibration robustness and generalisation. This automated optimisation method significantly improves the accuracy and efficiency of ASM calibration, advancing intelligent wastewater process modelling.

This study presents a multi-agent reinforcement learning (MARL) framework for integrated real-time control (RTC) of urban drainage systems (UDSs), coordinating sewers, wastewater treatment plants (WWTPs), and receiving waters. Trained within a hydraulic–water quality simulation environment using QMIX, the framework enables facility-level decision-making and adaptive system coordination. Applied to Lu’an City, China, MARL achieved a 25.4% reduction in flooding and overflow volumes and an 18.0% decrease in river pollutants relative to benchmark strategies, while maintaining real-time control feasibility (6.35 s per 5-min interval). Under rainfall forecast and sensor noise uncertainty, MARL improved performance stability by 44.7–52.4%. Despite operational trade-offs, the framework supports integrated system optimization and consistent water quality improvements in urban settings.

Machine learning-based non-target analysis (ML-based NTA) faces the critical challenge of linking complex chemical signals to contamination sources. This review proposes a systematic framework of ML-assisted NTA for contaminant source identification, emphasizing the strategies and considerations of key steps in data processing, pattern recognition, and model validation. The framework provides practical guidance for translating raw NTA data to actionable environmental insights that support informed decision-making.

Ternary heteronanocomposites Bi₂S₃/BiOCl@In₂S₃ and Bi₂S₃/BiOBr@In₂S₃ were designed as potent and sustainable photocatalysts and demonstrated their visible light-driven detoxification of the organoarsenic pollutant (carbarsone). A facile wet chemical-based synthesis method was applied to fabricate ternary nanocomposites, and their structural properties were analyzed using several analytical techniques (FETEM, XRD, XPS, etc.). In both nanocomposites, the monodispersed nanorod-structured Bi₂S₃ and nanosheet-structured In₂S₃, BiOCl, and BiOBr materials were combined to form multi-structured nanocomposites. Both nanocomposites exhibited good photocurrent responses and lower band gap energies that led to their use as photocatalysts for the degradation of carbarsone. As expected, Bi₂S₃/BiOCl@In₂S₃ and Bi₂S₃/BiOBr@In₂S₃ displayed excellent catalytic performance, achieving carbarsone degradation within 34 and 20 min, with corresponding higher rate constants of 0.1022 and 0.1572 min⁻¹, respectively. This enhanced photocatalytic activity arose due to synergistic double Z-scheme heterojunctions originating based on the band energies within nanocomposites, which can increase the inhibition of the photogenerated electrons and holes pair’s recombination and relatively maintain the strong catalytic redox properties. Bi₂S₃ acts as an interfacial mediator for effective charge separation, whereas In₂S₃ and BiOCl/BiOBr feasibly generate hydroxyl radical reactive species. Mass spectral analysis was employed to prove the oxidative pathway mechanism wherein hydroxyl radicals effectively degrade carbarsone. Furthermore, these nanocomposites displayed strong structural stability along with sustaining catalytic performance, and insignificant loss during the recycling processes. The integration of visible-light sensitivity with strong oxidative capabilities establishes Bi₂S₃, In₂S₃, BiOCl, and BiOBr as excellent candidates for photocatalytic detoxification of pollutants.

Pharmaceutical contaminants in water pose a significant ecological risk and require advanced treatment solutions. In this study, the strategy of oxygen-deficient TiO_2-x and doping is evaluated for its photocatalytic efficiency under visible light for pharmaceutical pollutant degradation, seawater purification and green fuel production. The oxygen vacancies reduced the band gap and improved visible light absorption and charge separation, allowing TiO_2-x to achieve almost complete degradation of aspirin within 6 h. Mechanistic studies (EPR, LC-MS) revealed •O₂⁻ and h⁺ as the dominant reactive species. The TiO_2-x (1:1) catalyst showed excellent stability and reusability. Modified catalysts (TiO₂-Cu, TiO₂-GO) were also evaluated, with TiO₂-Cu and TiO_2-x (1:1) showing superior removal of organic pollutants (>90%), natural organic matter (NOM) and divalent ions (Mg²⁺, Ca²⁺) in seawater. While efficient degradation reduced biotoxicity (95% EC50 reduction in the Microtox assay), incomplete mineralization in some systems resulted in toxic intermediates, highlighting the need for combined chemical and toxicity assessments. In addition, TiO_2-x (1:1) and TiO₂-GO showed increased activity in CO₂ reduction. This work highlights oxygen vacancy engineering as a promising strategy for visible-light-driven environmental photocatalysis.

Real-time modeling is vital for the intelligent management of urban water distribution systems (WDSs), enabling proactive decision-making, rapid anomaly detection, and efficient operational control. In comparison with traditional mechanistic simulators, data-driven models offer faster computation and reduced calibration demands, making them more suitable for real-time applications. However, existing models often accumulate long-term prediction errors and fail to capture the strong temporal dependencies in measured time series. To address these challenges, this study proposes the Kolmogorov–Arnold Attention Network for the real-time modeling of WDSs (KANSA), which combines Kolmogorov–Arnold Networks with attention mechanisms to extract temporal dependency features through bidirectional spatiotemporal processing. Additionally, a multi-equation soft-constraint formulation embeds mass and energy conservation laws into the loss function, mitigating cumulative errors and enhancing physical consistency. Evaluations on a benchmark network and a real-world system demonstrate that KANSA achieves high-accuracy real-time estimation and pattern fidelity while maintaining engineering-grade hydraulic balance.

Data-driven models often neglect the underlying physical principles, limiting generalization capabilities in water distribution systems (WDSs). This study presents a novel spatio-temporal graph physics-informed neural network (ST-GPINN) for water quality prediction in WDSs, integrating hydraulic simulations, physics-informed neural networks (PINNs), and graph neural networks (GNNs) to capture dynamics and graph-based network connectivity while approximating partial differential equations (PDEs). ST-GPINN discretizes WDSs using virtual nodes to enhance spatial granularity, employs an Encoder-Processor-Decoder architecture for predictions. Validated on Network A (a small-scale network with 9 junctions and 11 pipes) and Network B (a real large-scale WDS with 920 junctions and 1032 pipes), ST-GPINN outperforms others, achieving a MAE of 0.0073 mg/L, RMSE of 0.0121 mg/L, and R² of 88.91% in Network A, and a MAE of 0.008 mg/L, RMSE of 0.0098 mg/L, and R² of 98.91% in Network B. Its scalability and accuracy highlight ST-GPINN’s potential for water quality predictions.