2. 2
TRILITY
Operational sites and people snapshot
2
3
New South
Wales
7
8
Queensland
4
7
Victori
a
4
Tasmania
9
4
South
Australia
3
6
Western
Australia
2
7
New
Zealand
0
Australian
Capital
Territory
3. 3
3
Sustain
optimal
performance
Prescribe
best course
of action
Detect
changes
automatically
DECISIONS MADE CLEAR
Sustainable operational
excellence Change in conditions New process settings
required
Provide update from KDX
application
Automatically identify
underperforming processes
or assets
Digital tool recommends
the best course of action
Update control strategies
dynamically while reducing
costs and risk
4. 4
4
Decisions made clear
Defining a
new digital
landscape
We view utility
operations as a whole,
providing the
confidence to solve
common problems with
process twins and
automation.
Source water
quality and
climate
resilience
Effective water treatment
processes that optimize
chemical usage, filter
efficiency, and WQ compliance
management.
Demand forecasting, network
leakage detection and water age
management
Resilience to water sources
impacted by climate change
and emerging contaminants.
Wastewater resource recovery,
optimising nutrient removal and
reducing greenhouse gas
emissions.
Water
treatment
and
production
Wastewater
treatment
sustainability
Networks,
water quality,
and
customer
satisfaction
5. 5
Data quality and management
Influent quality and loading variability
Operator capacity and skills
Energy consumption costs
Residuals
The increasing complexity of managing
wastewater treatment
DECISIONS MADE CLEAR
6. 6
Data quality and management
The increasing complexity of managing
wastewater treatment
DECISIONS MADE CLEAR
⢠Data storage and retrieval
required for billions of
timeseries data points.
⢠Information and cybersecurity
costs and complexity for
distributed and cloud-based
systems.
⢠Aging solutions lacking
modern user interfaces and
experiences for ease of use
7. 7
Influent quality and loading variability
The increasing complexity of managing
wastewater treatment
DECISIONS MADE CLEAR
⢠WWTP capacity issues with growing
urban populations.
⢠Operators under resourced to identify
and adapt to toxic shock events.
⢠Lack of regular or comprehensive
influent WQ monitoring programs.
8. 8
Operator capacity and skills
The increasing complexity of managing
wastewater treatment
DECISIONS MADE CLEAR
â˘Tighter regulations on performance and
effluent quality.
â˘Age of assets and increased maintenance.
â˘Utility skills retention with an aging
workforce.
â˘Additional pressure to use more and more
systems and processes, often with limited
support.
9. 9
Energy consumption costs
The increasing complexity of managing
wastewater treatment
DECISIONS MADE CLEAR
⢠The water sector accounts for 4%
of total energy consumption, with
highly energy-dependent
wastewater treatment plants
(WWTPs) accounting for 25% of the
total energy use
⢠Globally, almost 400 billion m3 of
wastewater is produced annually,
and it is expected to increase by 25
and 50% by 2030 and 2050,
respectively.
10. 10
Energy consumption costs and net zero
The increasing complexity of managing
wastewater treatment
DECISIONS MADE CLEAR
⢠WWTPs in the USA generated 20
million metric tons (MMT) CO2-eq
in 2017.
⢠The energy needed for a typical
domestic WWTP employing aerobic
activated sludge processes and
anaerobic digestion (AD) is 0.6 kWh
per m3 of wastewater treated
⢠About half of which is used for
providing electrical energy to
sustain aeration basins.
11. 11
Residuals management
The increasing complexity of managing
wastewater treatment
DECISIONS MADE CLEAR
⢠Increased transportation and
disposal costs.
⢠Increased biosolids management
regulations and requirements.
⢠Increasing industry movement to
resource recovery and a need for
optimised production and recovery
techniques and solutions.
12. 12
Bioreactor health
DECISIONS MADE CLEAR
How do we balance optimal bioreactor health
energy use and operator time with changing
conditions and goals?
Is there a better way to think about bioreactor
health and managing process complexity?
13. 13
What options do we have?
DECISIONS MADE CLEAR
Offline PBSE models
Additional instrumentation
Training and development of staff
Advanced energy management
Realtime DPTs and machine learning
Capital upgrades
14. 14
14
Why now for digital process twins (DPTs)?
⢠No or minimal retroffiting of existing plant.
⢠New powerful mathematical optimisations and solvers
have been developed.
⢠ML capable of real-time process optimisation has
advanced rapidly in the last 10 years.
⢠Numerical simulation performance and vastly
increased computing power and efficiency.
⢠Improved economics with reduced cost of ownership
and increased availability.
DECISIONS MADE CLEAR
16. 16
KDX operations platform
⢠Provide a cost effective and
scalable DPT platform.
⢠Decision support tools co-designed
with operations teams.
⢠Designed from the ground up to
meet information security needs for
utility infrastructure.
DECISIONS MADE CLEAR
17. 17
17
Victor harbor WWTP
⢠Operated by TRILITY since 2006
⢠Proof of concept plant for a scaled dynamic
co-simulation environment 5MLD
Membrane
bioreactor
WWTP
DECISIONS MADE CLEAR
Units 50th percentile 90th percentile
BOD5 mg/L 5 10
SS mg/L 1 2
NH3-N mg/L 1 2
TN mg/L 5 10
TP mg/L 0.05 0.1
E. Coli E.c/100ml 1 10
18. 18
Real-time co-simulation
DECISIONS MADE CLEAR
ODE + ML
Surrogate
Calibration and validation campaign
with dynamic WWTP real-time data
ingestion
Physics based
simulation engine
Self-calibrating process
twin for the WWTP
Physics based
simulation engine
19. 19
Dynamic calibration of the DPT
⢠Multi-phase approach to
quantifying and reducing
uncertainty: long-term
averages down to
dynamic, near-term data.
⢠Calibration can be
adjusted based on real-
time and/or varying data
quality.
⢠Historical and real-time
âsoft-sensorsâ and insights
can then be generated.
DECISIONS MADE CLEAR
Multi-scale calibration to reduce uncertainty
22. 22
Benefits of real-time DPT calibration
⢠Soft sensing of Influent Water Quality and in-reactor âhiddenâ states.
⢠Early detection of toxic-shock or equipment failure.
⢠Long-term trending of parameters.
⢠Testing and validation of assumptions.
⢠Single- and/or multi-objective optimization.
24. 24
Physics-Informed Artificial Intelligence Systems
DECISIONS MADE CLEAR
⢠Pure physics relies heavily on quality
input data to predict ârealâ results, slow
to recalibrate, need expertise to build
and develop.
⢠Pure AI models need extensive
amounts of data in order to capture
full system dynamics.
⢠Surrogate models such as PINNs,
Bayesian ODEs etc can bridge this
gap.