• F E AT U R E
Key Takeaways
Water utilities are turning to
“digital twins” to benefit their
ongoing operations, improve
planning, and enhance
operator training.
As facility “flight simulators,”
digital twins allow utilities
to quickly investigate many
solutions without putting
equipment, public health, or
the environment at risk.
Dynamic process simulations
boost communication and
collaboration, especially for
teams with varied technical
backgrounds.
Digital Twins:
The Next Generation
of Water Treatment
Technology
Jason M. Curl, Tyler Nading, Kyle Hegger,
Amer Barhoumi, and Monika Smoczynski
44 JOURNAL AWWA
Layout imagery by pcruciatti, 13.20 Team, Mariusz Szczygiel/Shutterstock.com
DECEMBER 2019 • VOL.111 • NO.12 45
• F E AT U R E Wa te r U t ili t y S imul a t i o n s
W
ater and wastewater utilities made a While these dynamic models can be used for facilities rang-
technological leap forward with the ing in size from 50 gpm to hundreds of millions of gallons per
introduction of supervisory control and day, they provide the most value when there are complex in-
data acquisition (SCADA) systems. The terconnected systems or unique operational challenges that
next frontier of technological advancements is the use of would benefit from improved understanding.
facility “flight simulators,” or “digital twins,” that enable
dynamic process simulations. Many water utilities are Model Calibration
investigating or implementing digital twins to improve Digital twins can be created for both greenfield and
design of new facilities and to support operations staff existing facilities. Calibration for newly constructed
training, reduce risk, and optimize operations.
If a single picture is worth a thousand words, then
a digital twin of full-scale treatment and conveyance
facilities is worth a billion (or more). Utilities are begin-
ning to recognize that dynamic simulation models— If a single picture is worth a thousand
which visually integrate design and operations compo- words, then a digital twin of full-scale
nents such as process flow diagrams, SCADA screens,
and piping and instrumentation diagrams—can treatment and conveyance facilities
improve communication between project stakeholders, is worth a billion (or more).
especially for teams with varied technical backgrounds
and experience.
Society is just beginning to benefit from connected
digital systems, often referred to as the Internet of
Things (IoT). From refrigerators creating shopping lists facilities is not required; however, calibration for existing
to highways that vary tolls based on traffic loads, it infrastructure is an important component of the devel-
is only a matter of time until connected devices, data opment of a valid digital twin model. For new facilities,
analytics, and real-time machine learning are com- model calibration can occur once the facility is opera-
monplace in every industry, including water. Here, we tional and data can be collected.
describe how utilities are leveraging advanced software Field data are measured with online instrumenta-
tools to improve design and operations, and we high- tion located throughout the system—e.g., flowmeters,
light some of the lessons they learned. pressure indicators, and level measurements. However,
How does a digital twin work within the context of a historical data often have gaps from when the measuring
water treatment facility? Figure 1 graphically represents devices were not active, and there may be periods when
the flow of information among connected platforms, the devices provided erroneous data that do not reflect
which forms the foundation of a digital twin. The digital actual performance.
twin becomes the central clearinghouse for data and The first step in calibrating any model is to iden-
visualization because it accurately represents the full- tify events for comparison with the model results.
scale facility and its processes. Depending on the type of system that is modeled, a
minimum flow, average flow, and maximum flow event
Benefits of Digital Twins may be needed to properly calibrate the model. For
Model Characteristics each event that is selected, flow, pressure, and level
Digital twins can be leveraged to manage and operate all measurements are needed at as many points in the
types of water-related infrastructure, including pump system as possible.
stations, pipe networks, storage tanks, and treatment For an accurate digital twin, additional information
facilities. Digital twins for water treatment and convey- that helps simulate a dynamic event includes pump
ance systems commonly have several key characteristics. speeds, valve positions, gate positions, and data from
They typically any other equipment used to convey flow in the system.
••are hydraulically accurate for both pumped and grav- If the digital twin is capable of process simulation, then
ity systems, additional water quality and process performance data
••have control logic that mimics system operation, may be needed.
••operate on short (typically one-second) time steps, During calibration, the hydraulic characteristics of
••have an intuitive graphical user interface (GUI), and the model are adjusted until the flows, tank levels, and
••may include water quality and process performance. pipe pressures predicted by the model trend within a
46 JOURNAL AWWA
reasonable tolerance to what was observed during the intermediate pump station solely on the basis of feed
event. Typical hydraulic characteristics that require wet-well level, a digital twin can inform an improved
adjustment include pump performance curves, orifice control approach; this could include a feed forward
discharge coefficients, and pipe roughness. As addition- flow, from a summation of upstream flowmeters, plus
al events are tested and hydraulic characteristics are a wet-well-level secondary trim. Additionally, pump
adjusted, previous test events can be resimulated to ver- transitions could be improved to minimize power de-
ify how the adjustments affect the model’s predictions mands or starts.
over the range of flows. Ultimately, the use of digital twins promotes better
The process of model calibration involves in-depth overall system understanding, which also improves
holistic evaluation of whole-plant
data sets. While evaluating these
large data sets, modelers can
commonly identify characteristics
or diagnose underperforming fa-
cilities that may not be completely
understood by facility staff. Faulty A Digital Twin Forms a Central
or improperly scaled instrumenta- Repository for Information and
tion, degraded pump performance,
and hydraulic bottlenecks are
Provides a Basis for Analysis for
examples of such findings. This Treatment Facilities
is an indirect result of model cal-
ibration, which is useful for older
facilities that have been retrofitted
multiple times with varying levels
of engineering and documentation.
Facility Commissioning
When starting up a new facility, Database
Predictive
of Flow
one that has undergone retro- Event
Events
Analysis
fit construction, or an existing
facility that is upgrading its con-
trol system, use of a digital twin
greatly reduces risk in the early Hydraulic Model
days of operation. The ability to Calibration with Operator
test different control approaches Actual Plant Data Training
in a virtual environment allows
for rapid analysis of many alter-
natives without risking damage Dynamic Process
to equipment, overflows, or poor Modeling
quality of treated water. In turn,
this encourages creative thinking
to identify new approaches that Optimization,
Machine Learning,
may deliver better solutions. and Data Reduced Risk and Improved
The ability to test new con- Analytics System Understanding
trol approaches combined with During Design
increased understanding of the
effects of these changes over the
range of facility performance
results in improved process con-
trol design and identification of
control tuning parameters. For
example, instead of controlling an Figure 1
DECEMBER 2019 • VOL.111 • NO.12 47
• F E AT U R E Wa te r U t ili t y S imul a t i o n s
the ramifications of their decisions without risking
overflows or damaging equipment.
Failure Analysis
With a complete digital twin of a treatment facility,
detailed analyses can be performed for extreme cases
including failures. For example, a scenario for treat-
ing maximum design flows can be tested with major
equipment offline; likewise, a scenario for treating
minimum startup flows can be verified with equip-
ment turndown. Team members can increase their
comfort levels with operating under these extreme
conditions in the safety of the simulated environment.
An added benefit of testing extremes is identification
of potential shortfalls in the system’s capabilities or re-
alization of additional unplanned capacity. Without a
computer simulation model, these conditions can only
be observed under true duress, which does not allow
for careful analysis and study.
Analytics, Machine Learning, and Predicting
Performance
The addition of data analytics to a digital twin can
A digital twin can be used for operator training and supervisory further increase its value for facility operations. Data
control and data acquisition system testing. streams from online instrumentation to laboratory
sample results can be centrally collected to identify
outliers, which provides an automated method for data
quality control. Once the outliers are identified and
facility commissioning as various low-flow and high-flow scrubbed, any data gaps can be automatically infilled.
cases can be tested in the computer environment well SCADA data can be backfilled once outliers are re-
before construction is completed. moved. Laboratory data can be infilled using season-
ality models, autocorrelation, and cross-correlation to
Operator Training interpolate between points.
A digital twin can be further leveraged as a powerful Once the data streams are analyzed and scrubbed,
operator training tool. With the GUI customized for they can be uploaded into the digital twin and simula-
easy navigation, the model can be used to isolate and tions can be run automatically. Parallel simulations can
explore specific layers, including an overall process flow also be run with data variations to understand the sensi-
diagram view, a detailed unit process view, an individ- tivity of the facility operations to the inputs. With results
ual equipment view, and a controls view. Data can be from the automatic simulations, facility performance for
plotted over time, and equipment performance can be the coming hour, two hours, or even multiple days can be
animated for items such as pump speeds, valve posi-
tions, water levels, pressures, flows, and more. The dig-
ital twin can even incorporate the various alarms and
notifications that operators would experience during
full-scale facility operation. Digital twins can be leveraged
The biggest benefit of using the digital twin as an to manage and operate all types
operator training tool is the ability to “operate” the
facility in conditions that are rarely observed in a safe
of water-related infrastructure,
environment. These conditions include equipment including pump stations, pipe
failures, extreme high or low flows, emergency shut-
downs, and system restarts. Additionally, operators
networks, storage tanks, and
can test their own operational theories to understand treatment facilities.
48 JOURNAL AWWA
An Example of the Digital Twin for San Diego’s North City Pure
Water Facility
Figure 2
predicted. On the basis of these predictions, operations quality. When the simulations are parallelized and con-
staff can be informed as to how to adjust their approach- nected to optimization protocols, multiple simulations
es to improve performance and/or reduce operational can be run—thereby increasing operators’ understanding
risk. Over time, the data analytics and performance pre- of the proper chemical doses to improve water quality as
dictions of the digital twin will improve, adding further well as to minimize production of solids and use of chem-
value to the facility operations. icals and power.
Case Study
Optimization of Facility Processes The City of San Diego is in the first stages of implement-
With a complete model, optimization can be conduct- ing the Pure Water San Diego program, which will supply
ed offline from the facility and could include linear and one-third of the city’s water by 2035. Phase one consists
nonlinear optimization approaches—e.g., the evaluation of multiple facilities, including new wastewater convey-
of pump energy efficiency and relationships to operation- ance infrastructure (Morena Pump Station and pipeline),
al requirements. Factors such as wet-well sizes, individual upgrades to the existing North City Water Reclamation
pump capacities, duty/standby requirements, actual Plant (WRP), the new North City Pure Water Facility
pump curves, constant or variable speed, sequencing, (PWF), and new purified water conveyance facilities to
ramp up, and ramp down can all be modeled and evaluat- Miramar Reservoir. A digital twin of the advanced water
ed using these powerful models. purification facility (Figure 2) will be used to improve
Another example is optimization of chemical dosing in commissioning and long-term operations as an operator
a treatment facility. With available online data incoming training platform.
to the simulation coupled with expected laboratory re- The North City PWF will accept treated tertiary
sults (from seasonality assumptions and previous data), effluent from the North City WRP. The PWF employs
the digital twin can simulate expected finished water ozone, biologically active carbon (BAC) filtration,
DECEMBER 2019 • VOL.111 • NO.12 49
• F E AT U R E Wa te r U t ili t y S imul a t i o n s
microfiltration (MF), reverse osmosis (RO), and optimize their facilities offline under “normal” condi-
ultraviolet–advanced oxidation process treatment tions and prepare for worst-case scenarios.
systems. This facility accepts more than 40 mgd of Testing what-if scenarios using these calibrated models
treated effluent and will produce up to 30 mgd for is straightforward and helpful. Additionally, digital twins
downstream potable water supply. enable control systems to be tuned before starting oper-
The city is employing a digital twin of the PWF to ations, which can significantly reduce startup time and
reduce potential future operational challenges. This risk. Because the model can be run over the range of de-
simulation model runs on one second time steps and sign conditions, control parameter tuning is more robust.
fully replicates the facility’s hydraulics, controls, As the world becomes more connected through the
and process performance. The dynamic nature of IoT, water treatment technology and performance will
the model allows analysis beyond simply considering become more reliant on data analytics and dynamic pro-
a minimum, average, and maximum f low case. The cess simulation. Moving forward, more utilities will use
periodic nature of backwashes, f lushes, drains, and digital twins to continuously improve operations with
recycles can be modeled. respect to solids production, chemicals, and energy—all
The digital twin can also mimic the control schemes while ensuring high-quality treated water is delivered to
for the PWF. The BAC process feeds a wet well before their customers. The use of digital twins is the next step
pumping to MF. Additionally, there is a feed wet well in the future of water treatment.
before RO. These intermediate wet wells present risks
for having unit process operations out of sync. The About the Authors
model can be used to verify control points as to when
Jason M. Curl is a principal water treat-
pumps should start and stop and how they should adjust
ment technologist and digital twins
speed in relation to other process operations and online
global technology leader at Jacobs in
instrumentation.
Englewood, Colo. (www.jacobs.com);
For example, BAC backwash operations will routinely
[email protected].
affect the volume of water in the MF feed tank, which is
meant to serve consistent flow to downstream MF and Tyler Nading is a drinking water and reuse technologist
RO systems. Before construction, proactive and reac- at Jacobs, Englewood, Colo.
tive control strategies can be tested in the digital twin
under various scenarios to mitigate impacts of volume Kyle Hegger is a digital twin project technologist at
fluctuations. Jacobs, San Diego, Calif.
Finally, the dynamic simulation model incorpo-
rates predictive process modeling. Water chemis- Amer Barhoumi is lead project manager at City of San
try throughout the facility can be simulated so that Diego, San Diego, Calif.
chemical feed rates and use can be predicted as they
relate to meeting water quality goals. Water treatment Monika Smoczynski is assistant project manager at City
can be simulated using variable feedwater quality to of San Diego, San Diego, Calif.
leverage the dynamic nature of the model running over
extended periods. Another key water quality param- https://doi.org/10.1002/awwa.1413
eter that can be monitored is process performance,
such as RO membrane scale potential, finished water
stability indexes, and log removal value of various AWWA Resources
microorganisms. •• Consider the Value of Hydraulic Modeling for Operators.
Huang, M., 2019. Opflow, 45:7:16. https://doi.org/10.1002/
The city’s operators will also use this powerful com-
opfl.1216.
puter simulation for training and to increase their under- •• Evaluating Water Distribution System Response to
standing of the advanced water purification treatment Emergency Events. Chenevey, B., 2019. Journal AWWA,
processes onsite. 111:6:32. https://doi.org/10.1002/awwa.1304.
•• Risk and Resilience Assessment Isn’t Optional Anymore.
Walski, M., 2019. Opflow, 45:9:22. https://doi.org/10.1002/
Conclusions opfl.1249.
Facility digital twins that include dynamic simulations
for hydraulics, controls, and water quality can provide These resources have been supplied by Journal AWWA staff.
For information on these and other AWWA resources, visit
an interactive, full-motion picture of a facility before
www.awwa.org.
it’s built. With a flight simulator model, operators can
50 JOURNAL AWWA
