Hybrid Models in Wastewater: Merging Mechanistic with Data-Driven

On one hand, mechanistic models have been extensively used for process design and optimization. However, building the perfect mechanistic model is only possible for the simplest systems, consequently not for biological processes. The resulting limited prediction capability is a limiting factor in using them for real-time decision-making. On the other hand, “pure” big data models exist but remain isolated initiatives in the academic sector.

The close integration of mechanistic models with Big Data processing and learning techniques has the potential to transform how the day-to-day integrated operation of UWS is currently being done.

In this webinar, we will showcase the potential of hybridized mechanistic and data-driven models in the wastewater sector by presenting three existing initiatives that are working towards integrating both paradigms into hybrid models.

Participants are expected to have basic knowledge of: water and wastewater treatment fundamentals and process modelling, data collection and reconciliation in the water sector, and data science.

Panelists

• Lina Belia https://iwa-connect.org/user/5660ad04340f2a9d1260c5b7 – Primodal Inc., Canada
• Ulf Jeppsson https://iwa-connect.org/user/566823e1c82261aa7e3903bf – Lund University, Sweden
• Kris Villez https://iwa-connect.org/user/56682481c82261aa7e3913a8 – Oak Ridge National Laboratory, USA
• Feiyi Li https://iwa-connect.org/user/5c08e935f3f2f469d389b2e7 – modelEAU, Canada
• Junjie Zhu https://iwa-connect.org/user/593ade2da1e0c73b232b1aca – Princeton University, USA
• Ward Quaghebeur https://iwa-connect.org/user/5e1485ac66488f47d31c43c1 – BIOMATH - Ghent University, Belgium
• Bram De Jaegher https://iwa-connect.org/user/5e4e5ac027bc6660856ea778 – BIOMATH - Ghent University, Belgium

process modelers; utility managers; technologists, consultants, academics and modelers with an interest in data science

Following this webinar, participants will:

● Understand the advantages and drawbacks of mechanistic and data-driven models, and that neither can tackle all the modelling challenges of the future. ● Be familiar with model hybridization through practical examples, combining mechanistic and data-driven models to leverage the strength of both approaches.