(13f) Hybrid Modeling of Wastewater Treatment Plants for Greenhouse Gas Emissions Prediction and Control

Wastewater treatment plants (WWTPs) and water resource recovery facilities (WRRFs) play a critical role in environmental protection, but they also contribute to greenhouse gas (GHG) emissions through both direct and indirect ways. Direct emissions are related to the underlying bioprocesses, while indirect emissions are related to energy consumption. Therefore, predicting and controlling GHG emissions in WWTPs/WRRFs is crucial for mitigating environmental impact and meeting regulatory standards. However, predicting and controlling these emissions is difficult due to the complex, nonlinear nature of the biological processes within WWTP/WRRFs, as they are highly influenced by numerous operational and environmental factors [2]. Addressing these challenges requires advanced modeling approaches capable of improving predictive accuracy.

Significant efforts have been made to create accurate first-principles mechanistic models for WWTPs/WRRFs. For example, the family of activated sludge models (ASMs) that capture the biochemical processes underlying wastewater treatment are used and studied in the literature extensively [3–5]. However, these models often struggle with predictive accuracy for real-time optimization due to inherent process uncertainties or other underlying mechanisms that are unaccounted for [6]. To enhance model predictive capabilities, data-driven approaches have been explored, leveraging machine learning techniques to improve predictions [7,8]. Despite their promise, pure data-driven models often lack interpretability, generalizability, extrapolability, and adherence to the physical laws [9,10]. Therefore, in recent years, there has been some preliminary investigation into models that integrate both mechanistic and data-driven components [1,6,11,12]; however, there is still much research to be done on the architectures of the integration as well as the application to specific WWTP configurations.

To address these gaps, we propose several hybrid modeling approaches for developing highly predictive digital twins of WWTPs, with a specific interest in GHG reduction. Our case study utilizes data from the University of Connecticut’s WWTP, focusing on biological nutrient removal (BNR) and secondary clarification. The mechanistic model is based on the ASMN_G biokinetic framework [13], while the data-driven component refines predictions by leveraging machine learning techniques. By integrating these hybrid models within an economic model predictive control (EMPC) framework, we enable dynamic optimization of treatment operations to minimize GHG emissions while maintaining energy efficiency and effluent quality. This approach improves real-time decision-making and enhances the sustainability of WWTPs.

In this presentation, we will discuss the development and validation of the proposed hybrid models, highlighting their advantages over traditional approaches. We will demonstrate how hybrid models can be used to optimize WWTP operations, showcasing the impact of hybrid modeling within an EMPC framework. Additionally, we will explore the trade-offs, computational challenges, and potential future improvements in GHG emission control strategies. Through this discussion, we aim to provide insights into the role of hybrid modeling in advancing sustainable wastewater treatment.

References:

1. Khalil M, AlSayed A, Elsayed A, Sherif Zaghloul M, Bell KY, Al-Omari A, et al. Advances in GHG emissions modelling for WRRFs: From State-of-the-Art methods to Full-Scale applications. Chem. Eng. J. 2024;494:153053.
2. Lancioni N, Szelag B, Sgroi M, Barbusiński K, Fatone F, Eusebi AL. Novel extended hybrid tool for real time control and practically support decisions to reduce GHG emissions in full scale wastewater treatment plants. J. Environ. Manage. 2024;365:121502.
3. Gujer W, Henze M, Mino T, Loosdrecht M van. Activated sludge model No. 3. Water Sci. Technol. 1999;39:183–93.
4. Blomberg K, Kosse P, Mikola A, Kuokkanen A, Fred T, Heinonen M, et al. Development of an Extended ASM3 Model for Predicting the Nitrous Oxide Emissions in a Full-Scale Wastewater Treatment Plant. Environ. Sci. Technol. 2018;52:5803–11.
5. Corominas L, Flores-Alsina X, Snip L, Vanrolleghem PA. Comparison of different modeling approaches to better evaluate greenhouse gas emissions from whole wastewater treatment plants. Biotechnol. Bioeng. 2012;109:2854–63.
6. Quaghebeur W, Torfs E, De Baets B, Nopens I. Hybrid differential equations: Integrating mechanistic and data-driven techniques for modelling of water systems. Water Res. 2022;213:118166.
7. Khalil M, AlSayed A, Liu Y, Vanrolleghem PA. Machine learning for modeling N2O emissions from wastewater treatment plants: Aligning model performance, complexity, and interpretability. Water Res. 2023;245:120667.
8. Vasilaki V, Danishvar S, Mousavi A, Katsou E. Data-driven versus conventional N2O EF quantification methods in wastewater; how can we quantify reliable annual EFs? Comput. Chem. Eng. 2020;141:106997.
9. Li K, Duan H, Liu L, Qiu R, van den Akker B, Ni BJ, et al. An Integrated First Principal and Deep Learning Approach for Modeling Nitrous Oxide Emissions from Wastewater Treatment Plants. Environ. Sci. Technol. 2022;56:2816–26.
10. Michałowska K, Goswami S, Karniadakis GE, Riemer-Sørensen S. Neural Operator Learning for Long-Time Integration in Dynamical Systems with Recurrent Neural Networks [Internet]. In: 2024 International Joint Conference on Neural Networks (IJCNN). 2024 [cited 2025 Apr 2]. page 1–8.Available from: https://ieeexplore.ieee.org/document/10650331/citations?tabFilter=paper…
11. Schneider MY, Quaghebeur W, Borzooei S, Froemelt A, Li F, Saagi R, et al. Hybrid modelling of water resource recovery facilities: status and opportunities. Water Sci. Technol. 2022;85:2503–24.
12. Mehrani MJ, Bagherzadeh F, Zheng M, Kowal P, Sobotka D, Mąkinia J. Application of a hybrid mechanistic/machine learning model for prediction of nitrous oxide (N2O) production in a nitrifying sequencing batch reactor. Process Saf. Environ. Prot. 2022;162:1015–24.
13. Kim D, Bowen JD, Kinnear D. Comprehensive Numerical Modeling of Greenhouse Gas Emissions from Water Resource Recovery Facilities. Water Environ. Res. 2015;87:1955–69.