Slide 1: Digital Twins: Enabling Data-Driven Water Reclamation/Reuse Solutions
Digital Twins: Enabling Data- Data-Driven Water Reclamation/Reuse Solutions In the kNOW Webinar Series August 26, 2020
Slide 2: Safety Moment – Handling home deliveries
Safety Moment – Handling home deliveries
Slide 3: What is a Digital Twin?
What is a Digital Twin? A digital representation of a physical system coupled with real-time data, that can be used for synthetic data generation, scenario analysis, performance prediction and operational optimization
Slide 4: Relevant Terminology
Relevant Terminology • Digital Replica • A model without autonomous data exchange with physical system • Digital Twin • A Digital Replica with autonomous information exchange to/from physical system
Slide 5: Agenda
Agenda • Simulating the Big Picture: An Integrated System Modeling Approach for Operational Resiliency – Garrett Owens, Jacobs • Improving Process Efficiency Through Better Data Acquisition and Visualization – Jim McQuarrie, MWRD • Utility Drivers and Experience with Data Driven Models – Adrienne Me
Slide 6: Simulating The Big Picture
Simulating The Big Picture An integrated system modeling approach for operational strategy Garrett Owens Digital Twins Global Technology Leader, Jacobs
Slide 7: Acknowledgments
Acknowledgments Monika Smoczynski • NCWRP Expansion Project Manager (City of San Diego) Mark Elliott • Client Manager (Jacobs) Troy Matsuura • Project Manager (Jacobs)
Slide 8: San Diego's Water Supply
San Diego’s Water Supply 85% of San Diego’s water is currently imported… ©Jacobs 2019
Slide 9: We Face Numerous Water Challenges
We Face Numerous Water Challenges Imported water cost per acre-foot Imported Water Cost Limited local & imported Natural disaster risk per Acre-Foot supplies Rising imported water costs Population growth Recurring drought Today Bay Delta constraints ©Jacobs 2020
Slide 10: San Diego's Water Supply System
San Diego’s Water Supply System Point Loma WWTP / Ocean Discharge Existing System… Primarily a Single-Use System Pure Water… Completes Our Water Cycle ©Jacobs 2020 10
Slide 11: Treatment & Storage Capacity / Pure Water process
RW Demand and Storage Capacity 50 40 Treatment & Storage Capacity Flow (MGD) 30 Constant Flow Demand 20 Recycled Water System 10 0 PS64 0 4 8 12 16 20 24 Water Quality Criteria … Miramar Reservoir Water Reclamation Plant Pure Water Facility Morena Pump Station • Influent Pumping • Ozone Raw Sewage F
Slide 12: Water Reclamation Plant Digital “Replica”
Water Reclamation Plant Digital “Replica” ©Jacobs 2020 12
Slide 13: Water Reclamation Plant Digital “Replica”
Water Reclamation Plant Digital “Replica” ©Jacobs 2020 13
Slide 14: Summary
Summary • Pure Water System is highly complex interaction between multiple facilities • Data manually imported for calibration of existing components and from EPANET to represent inflows • Digital Replica improves our understanding of the system and how to operate it • Used as a “flight simulator” f
Slide 15: Improving Process Efficiency through Better Data Acquisition and Visualization
Improving Process Efficiency through Better Data Acquisition and Visualization Jim McQuarrie Director of Planning and Innovation Metro-Wastewater Reclamation District 10 | oo 4 ° 20 40 cr) 80 To) 120 Applied CT Current k (Decay) North PAA Dosing [NORMGORIMONEANEN 7575 .200u Prediction Tool Influent
Slide 16: The COVID-19 crisis presents an opportunity that few feel equipped to pursue
The COVID-19 crisis presents an opportunity that few feel equipped to pursue. Although most executives agree that innovating the business will be critical ... 90% 85% O O are concerned that the believe that the COVID-19 crisis will COVID-19 crisis will have fundamentally change a lasting impact on t
Slide 17: Getting Back to Normal
Getting Back to Normal ¢ Continuity of Core Operations and Service ¢ Budget Cuts and Balanced Budgets ¢ Supply and supply chains
Slide 18: Metro Wastewater Reclamation District
Metro Wastewater Reclamation District * 2+ Million PE * Tiny Receiving Stream * Converted from Chloramine Disinfection to Peracetic Acid * Water Quality Benefits * Operational Simplicity | High Unit Price for PAA
Slide 19: Methods of modeling
Methods of modeling | MAVEN performance . } j @ -_, Online: Flow, temperature, SRT, nutrients, N N visual spectrum (2018) * Instant, short-term (5-60 min), 24-h avg Lab: 24-hour flow composite nutrients, TSS @ PAA & E. coli: 168 total sampling events in 2018 and 66 in 2019 Artificial neural networks
Slide 20: Disinfection Performance Dashboards
Noth Disinfection Perioemance: PrePAAE. cob © PostPAAE. coll» Ecol Limt og) oo North Disinfection Performance Evaluation | [seo over PAA Performance | [ Go to South Disinfection Performance Evaluation so so} a 30 Li Fe 5 g 4 r 7 o 1 oo | 0 Ps o Ps % 100 a opted CT Current k (Decay) - North PAA Dosin
Slide 21: Primary Treatment
Primary Treatment O71 Optimi
Slide 22: Primary Treatment NNM
Primary Treatment NNM Record Report ‘ Report for Neural Network Model Predict TSS Removal SOR.Smitn.NPRI.3.Flw Basic Summary: 2 Call: nnet.formula(formula = Optimization.NPRI.3.E_TSS_actual ~ © SOR.Simulation.NPRI.3.Flow + Optimization.NPRI.3.SOR, data = the.data, size = 10, linout = TRUE, rang = c(
Slide 23: Moving To a New Normal
Moving To a New Normal * Highly Integrated Municipal Services * Resource Sharing and Resource Efficiency * Data Analytics + Automated Supply Chain Management * Predictive Process and Maintenance Control * Massive Shift in Work Force Talent * Changing Expectations of the Workplace THIET:
Slide 24: Utility Drivers and Experience with Data Driven Models
Utility Drivers and Experience with Data Driven Models Adrienne Menniti, PhD, PE Principal Process Engineer, Clean Water Services
Slide 25: Who is Clean Water Services?
Who is Clean Water Services? • Mid-sized utility in suburban Portland, Oregon • 2 bigger facilities treating 25-35 mgd on average • 1 small facility treating 5 mgd on average • 1 small seasonal facility • Stringent seasonal phosphorus limit has been an innovation driver since the mid-1990s • Culture
Slide 26: Where does CWS see opportunities for data driven models or digital twins?
Where does CWS see opportunities for data driven models or digital twins? • Reduce human time spent sorting through data to support decisions • Make better decisions if the tool provides a more comprehensive view or provides decision support information in a more streamlined way But. . . • We always
Slide 27: What is hindering our adoption of the technology?
What is hindering our adoption of the technology? • Uncertainty with longevity of technology platforms • Relatively high cost to implement widely • Lack of experience with the technology - unknown benefits/unknown risks How to move forward? • We are looking for smaller ways to gain experience with a
Slide 28: Pilot Project Example
Pilot Project Example Data driven models to predict BPR stability
Slide 29: Driver: No tertiary coagulant = no BPR back up
Driver Stop using tertiary coagulant (alum) to meet our stringent phosphorus limit • Expecting new stringent limit on effluent aluminum (and iron) • Working with our regulatory agency to shift the format of our phosphorus permit to allow little or no tertiary coagulant use • Pilot this summer operat
Slide 30: CWS search for early warning of BPR upset
CWS search for early warning of BPR upset • If you are only monitoring secondary effluent, BPR will be “stable” for long periods of time only to get upset unexpectedly. • We have found ways to predict BPR upsets days or week in advance. Residual Phosphorus Uptake Batch Testing (RPU) P Uptake Value (
Slide 31: CWS search for early warning of BPR upset
CWS search for early warning of BPR upset We found a partner developing machine learning applications with an NSF grant Results with historical data looked promising • The combination of our innovation drive as an agency and the BPR application we could offer was serendipity • The grant offered a fe
Slide 32: CWS search for early warning of BPR upset
CWS search for early warning of BPR upset Preliminary results enough to convince NSF to provide an additional $1M grant to put the software online this summer through next summer.
Slide 33: What we've learned so far
What we’ve learned so far As a utility, we have a lot of “high quality” data already, which is what made the data driven model successful. • Our investment in data quality and infrastructure to support it (humans, instruments, lab) was very important. • The model needed a data stream that had a lot
Slide 34: Developing a “true” Digital Twin: The Changi WRP Story
Developing a “true” Digital Twin: The Changi WRP Story Bruce Johnson, PE, BCEE Wastewater Technology Fellow, Jacobs
Slide 35: Changi Water Reclamation Plant (CWRP) Singapore
Changi Water Reclamation Plant (CWRP) Singapore • Operated by Singapore Public Utilities Board (PUB) • Currently treating an average of approximately 920,000 m3/d (243 MGD) of used water • Fed by a deep-tunnel sewer system • Currently four bioreactor trains. Each train includes primary treatment and
Slide 36: Changi WRP Digital Twin Research Project
Changi WRP Digital Twin Research Project • A research collaboration between PUB and Jacobs in the development and application of a Digital Twin of CWRP • The Digital Twin includes models of: • Full plant Hydraulics (liquids and solids) • All major process controls • Biochemical Process • All impleme
Slide 37: CWRP Digital Twin Proposed Scope
CWRP Digital Twin Proposed Scope What does it take to develop a “True” Digital Twin: The Changi WRP Story ©Jacobs 2020
Slide 38: Overall Digital Twin Structure
Overall Digital Twin Structure Software Tools Functionality User Interface Realtime Data Machine Learning Integration, Analytics & to Improve Validation Functionality Custom Web Based Calibrated Integrated Interface Hydraulic and Hydraulic Control, Control Model Process Model Off-line operator Train
Slide 39: Changi WRP Process Flow Diagram Overview
Changi WRP Process Flow Diagram Overview NewWater (4) Stacked (8) Stacked 5-Pass Step-Feed (35 Primaries Secondaries CSTRs) Ocean Outfall MLE MBR 3 More Identical Trains Coarse Screening and Pumping Drying (10) WAS Thickening (10) Dewatering Centrifuges (5) Anaerobic Centrifuges Digesters (10) PSD T
Slide 40: Partial PFDs (one of 4 liquids modules + Solids)
Partial PFDs (one of 4 liquids modules + Solids)
Slide 41: CWRP Digital Twin Models
CWRP Digital Twin Models What does it take to develop a “True” Digital Twin: The Changi WRP Story ©Jacobs 2020
Slide 42: CWRP Replica Hydraulics and Control Model
CWRP Replica™ Hydraulics and Control Model
Slide 43: CWRP Sumo Process Model (169 biological reactors)
CWRP Sumo Process Model (169 biological reactors) ©Jacobs 2020 43
Slide 44: Live Data Make it a Digital Twin
Live Data Make it a Digital Twin • Outlier detection – identify anomalous data through various analytical methods • Infilling: Both on-line and laboratory bad/missing data • Process Deviations • Compare Model and cleaned actual data ©Jacobs 2020 44
Slide 45: Machine Learning Reduces Utility Effort
Machine Learning Reduces Utility Effort • One of the most time consuming tasks: keeping the model relevant • Machine learning auto- calibrates model, thus keeping it current • Requires direct connection to plant data ©Jacobs 2020 45
Slide 46: Predictive Analytics: Wastewater “Weather” Forecasting
Predictive Analytics: Wastewater “Weather” Forecasting • Use Monte Carlo tools to predict likelihood of events – 1, 3, and 5 days into the future – likelihood of clarifier overloading – high nitrates, etc. • Uses simplified model for speed – “Only” 40 bioreactors • Provide results in “Weather App” t
Slide 47: CWRP Digital Twin Example Results
CWRP Digital Twin Example Results What does it take to develop a “True” Digital Twin: The Changi WRP Story ©Jacobs 2020
Slide 48: Soft Sensors for Influent Quality
Soft Sensors for Influent Quality • A “soft-sensor” uses the model to estimate what a sensor might have read, without needing the sensor • Reduces capital and maintenance costs • CWRP Raw Sewage Characteristics (ammonia example) • Uses primary effluent ammonia measurements (on-line and laboratory)
Slide 49: Looking Forward: Moving from Reactive to Proactive Operations
Looking Forward: Moving from Reactive to Proactive Operations • Laboratory results are combined with predicted influent ammonia concentrations to develop 5-day dynamic predictions of COD concentrations ©Jacobs 2020 49
Slide 50: Digital Twin Summary
Digital Twin Summary Changi WRP Digital Twin replicates all significant aspects of a facility on a digital platform, Hydraulics, I&C, Process Benefits: • Increasing Productivity with: Real-time operation insights and process trouble-shooting. • Enhancing Resilience of Operations by: Moving from Reac
Slide 51: Acknowledgments
Acknowledgments • This research is supported by the National Research Foundation, Singapore, and PUB, Singapore’s National Water Agency under Urban Solutions & Sustainability (CRP(Water) RFP 1803 <PUB-1803-0014>) • This work could not have been accomplished without the extensive support, feedback, a
Slide 52: Digital Twins: Enabling Data-driven Water Reclamation/Reuse Solutions
Digital Twins: Enabling Data-driven Water Reclamation/Reuse Solutions • A digital representation of a physical system coupled with real-time data • Offers multiple benefits to the water industry • Development can be scaled to needs/capabilities • Degree of sophistication ultimately depends on level
Slide 53: Digital Twins: Enabling Data-driven Water Reclamation/Reuse Solutions – Q&A
Digital Twins: Enabling Data- Data-driven Water Reclamation/Reuse Solutions In the kNOW Webinar Series August 26, 2020 Q&A