1 of 13

SCIENTIFIC COLLABORATIONS

CLAIRE BEYELER & REGIS MOILLERON

Greater Paris Metropolis

session

2 of 13

For the public authority, scientific collaboration enables:

Sharing, collection and processing of data by a larger number of partners;
Criticism of public action and its direct impact on local environment and society;
Long-term perspective in strategy building, taking into account past events, global watershade perspective and potential impact of climate changes on drought and floods;
Education of representatives and populations on climate change, water, natural areas and their maintenance and restoration needs

20/03/2023

Water and Climate research priorities in greater Paris area

3 of 13

AI/ML Application for Wastewater Treatment Process�Optimization & Energy Demand Control

Two ongoing Case Studies of Scientific Collaboration

20/03/2023

Case studies of Scientific Collaboration

Paris (France)

Marcello SERRAO (W-SMART, LEESU & modelEAU, SIAAP)

Mocopée Research Program
W-SMART Research Projects
AI/ML Application: Innovative Tool for Process Optimization & Control
Twin Modeling Example
Results and Expected Outcome

New York (US)

Sam WHITE (W-SMART & NYU)

Industry needs for Energy Neutral & Demand Control
AI/ML Application for Smart Water Management
Application to Energy demand Control – example Nir Etsion (Israel)

4 of 13

20/03/2023

Collaborative Research Program in greater Paris area (SIAAP)

Mocopée Research Program

Modélisation, Contrôle et Optimisation des Procédés d’Épuration des Eaux

Operational Data in Real-Time

Process Expertise

Corporate Knowledge Base

University & Research Centres

Innovation Ressources

Industry

5 of 13

Artificial intelligence-based monitoring system & risk assessment method for geo-localization of leakage & early anomaly detection of water quality parameters of non-specific bio-contamination in water distribution systems (2019)
Lille University’s Smart Water Demonstration Site selected by EU for SW4EU Project

20/03/2023

Collaborative Research W-SMART

W-SMART Research & Collaboration Projects

Case studies France

Water Distribution System Analysis.
SVM’s and ANN’s identify anomalies and assess the severity-level.
Color-based GIS visualization of multi-spot risk analysis.

Tinelli, S., & Juran, I. (2019). Artificial intelligence-based monitoring system of water quality parameters for early detection of non-specific bio-contamination in water distribution systems. Water Supply, 19(6), 1785-1792.

Sponsors:

VITENS; Eau de Paris; Eaux du Nord;

EU-FP7 (10 M €)

6 of 13

Mocopée Research Program & W-SMART Participation

Goal: Improve Accuracy & Robustness of Process Model

Problem: Mechanistic models for biofiltration treatment are complex with many parameters that are difficult to calibrate & validate due to residual error or overfitting

Solution: Hybrid modelling�with an AI model to predict the mechanistic model errors

20/03/2023

Hybrid Model as an Innovative AI/ML Application for Process Optimization & Control

PhD Research Collaboration with LEESU, modelEAU, W-SMART and SIAAP (France) 2020 - 2023

Water and Climate research priorities in greater Paris area: Hybrid Biofilter Model

7 of 13

20/03/2023

Water and Climate research priorities in greater Paris area: Hybrid Biofilter Model

Wastewater Treatment Industry / Biofiltration

Primary Settling

Denit

Nit

Denit

River

Hybrid Model as an Innovative AI/ML Application for Process Optimization & Control

Case study Seine-Aval WWTP (1,3 million m3/day), Paris (France) managed by SIAAP

8 of 13

20/03/2023

Twin Model Structure

Mechanistic Model + Machine Learning Model = Higher Accuracy of Predictions

Water and Climate research priorities in greater Paris area: Hybrid Biofilter Model

9 of 13

20/03/2023

Results & Expected Outcome

Innovative AI/ML Application for Process Optimization & Control

[1] Stentoft, P.A., Munk-Nielsen, T., Vezzaro, L., Madsen, H., Mikkelsen, P. S., & Moller, J. K. (2019). Towards model predictive control: online predictions of ammonium and nitrate removal by using a stochastic ASM. Water Science and Technology.

[1]

Water and Climate research priorities in greater Paris area: Hybrid Biofilter Model

10 of 13

10

3/19/2023

WWTP Demand Response & Renewables

CAISO Net Load P(load) - P(rne): 2020 Trend

WWTP Energy Demand Profile

Water and Climate research priorities: WWT Smart Grid

Blue – Up Regulation Peaks (customers should consume more energy)

Red – Down Regulation Peaks (customers should consume less energy)

Left – WWTP energy demand profile one day, summer time

Right – CAISO net load: is defined as the total load on the power grid (i.e. power demand from consumers) minus input from variable generation resources (i.e. wind and solar). Therefore Power generated by conventional plants will become more variable. For instance, solar power comes in daylight hours, conventional generation doesn't need to produce around midday. Conventional PP require long periods of time to start or stop producing power, some cannot (i.e. nuclear, geothermal, hydro)

CAISO itself is not responsible for generating power for the grid, it oversees the energy production market and is charged with securing enough power to satisfy California's electricity demand at any given time.

11 of 13

Determine effect on quality parameters of Effluent and Sludge by Variable Demand Loads

Treatment Aeration

Wastewater Pumping

Sludge Pumping Transport

Support Auto-DR with AI assisted Intelligent Process Control and Optimization:

Pre-Processed Historical Plant Data
Statistical Mono and Multi-Parameter Analysis of COD, TSS and TKN
AI anomaly detection analysis in Auto-DR Scenario Simulations

Develop a WWTP Demand Response Optimization Tool for Energy-Water Systems Co-Optimization

Supports decision making for WWTP owners and operators
Energy System Operators gauge impact from water sector DR participation
Assists in cost/benefit analysis of plant upgrades to Energy System

Water and Climate research priorities: WWT Smart Grid

W-SMART Research & Collaboration Projects

Energy Neutral & Demand Control

Schematic of a WWTP

Highlighted equipment is suitable for DR

12 of 13

12

Database Processing

Functional Architecture

Statistical Analysis

Anomaly Detection System

Confidence level of anomaly likelihood
Severity of the anomaly
Risk of the anomaly
Sensor Deficiency

ASM

Models

Self-Corrective Action

Numerical Simulation

Database

AI Model Building

ASM/BIOWIN/GPS X

Effluent & Treatment Parameters

Long Short-Term Memory DNN

13 of 13

13

3/19/2023

Data Availability – Nir Etzion WWTP

The Nir Etizon WWTP is in the 400 Level of design & currently in the construction phase.

In order to ensure cost-minimization & 100% water reuse, the plant was designed w/ a digital-twin BIOWIN model integrated to an in-situ network of process sensors

Liquid line

Pretreatment – course and fine screen
Primary clarifier with possible fermentation
EQ tank
Biological treatment – A2O, MLE, UCT & RAS fermentation
Secondary clarifier
Deep bed sand filtration
UV and chlorine dioxide disinfection

Sludge line

Centrifuge thickeners
Anaerobic digesters for primary sludge
Aerobic digesters for WAS
Dewatering centrifuges

Average daily design flow: 32,000 m3/day (7.0 mgd)

Peak day design flow: 51,100 m3/day (11.2 mgd)

Influent

COD: 1,100 mg/l

BOD: 440 mg/l

TSS: 450 mg/l

TN: 80 mg/l

TP: 12 mg/l

Effluent

COD: 70 mg/l

BOD: 10 mg/l

TSS: 10 mg/l

TN: 10 mg/l

TP: 1.0 mg/l

E coli: 10 mg/l

Turbidity: 2.0 NTU

SAR: 5.0