Digital Twins
A digital twin is a live computer copy of a real thing, such as a machine or a power grid, kept up to date by sensors so you can test a change on the copy before touching the real one. Singapore built a twin of its national grid, with more than 18,000 transformers and 27,000 km of underground cable. Two years later, the part tracking transformer health was still a pilot. Building the model is easy; keeping it in step is the job.
Maturity
Where it feeds in
Where you have already met this
In 2021 Singapore's grid operator launched a digital twin of the national power grid - more than 18,000 transformers and 27,000 km of underground cable. Two years later, the part that tracks the health of those transformers was still being piloted. Building the model was the easy part. Keeping it in step is the job.
If this moved
If keeping twins in step got cheap, grid operators and factories could test changes before making them. Industrial hyper-automation, the built-environment deficit and precision everything would still sit at 2, held by managerial capability, digital transformation and smart cities; the energy transition sits at 1 on consumption norms.
What to watch for
Operators reporting twins still in use years after commissioning, rather than launches. Published checks of a twin's predictions against what the asset actually did.
Not to be confused with
A 3D or BIM model. A model stays as it was drawn. A twin is fed live data from the real thing - without the feed it is only a simulation.
What it is
Live computational models of a physical asset or system, kept in sync with it by sensor data and used to test changes before making them.
How it works
- 01Instrument the assetSensors on the transformer, turbine or building measure temperature, vibration, load and voltage.
- 02Stream the readingsGateways send the data continuously to the system that runs the model.
- 03Keep the model in stepThe model updates its state from the readings, so it describes the asset as it is today, not as it was designed.
- 04Ask 'what if'Engineers run simulations on the synchronised model: what happens if this line carries more load, if this pump fails, if a charger is connected here.
- 05Act - and keep checkingDecisions, sometimes automatic, go back to the real asset. The model has to be re-validated as the asset ages; a twin that drifts out of step keeps being trusted.
State of play
- SP Group launched a digital twin of Singapore's national power grid, covering more than 18,000 transformers, 11,000 substations and 27,000 km of underground cables (SP Group, 27 Oct 2021).
- SP Group's digital asset twin was still being piloted in October 2023 (Singapore International Energy Week, 24 Oct 2023).
- The Cordova, Alaska microgrid ran a real-time twin fed by one-second SCADA data, grid sensors and about 2,000 smart meters, as a research demonstration (IEEE Smart Grid bulletin, NREL and Cordova Electric Cooperative, Oct 2022).
- The US National Academies concluded that publicity around digital twins outweighs the evidence base of success (National Academies, Foundational Research Gaps and Future Directions for Digital Twins, 2024).
- A review screening 433 articles on manufacturing twins found no consistent, universally accepted definition of the term (Villegas, Macchi & Polenghi, Annual Reviews in Control 60, 2025).
Why it matters
It lowers the cost of finding out what a change would do. That is directly the Audit's fourth step - asking where the next unit of investment actually helps - made computable for physical systems.
- It lowers the cost of finding out what a change would do - the Audit's fourth step, made computable for physical systems.
- The model is cheap; synchronisation is what costs. That is why twins cluster where assets are instrumented anyway.
- It feeds industrial hyper-automation, the built-environment deficit and precision everything, all at 2. At 2 it joins the enablers that share those readings.
What holds it back
Capital for sensors and links
Dense sensors and reliable data links cost money up front and for the life of the asset, which confines twins to high-value infrastructure.
Upkeep
Someone has to keep the model matched to a physical thing that changes. That is a standing cost, usually budgeted as a one-off project.
Validation
Few twins publish how well their predictions matched reality, so buyers cannot tell a working twin from a demonstration.
Standards
Design, construction and maintenance data sit in incompatible formats, so a twin rarely survives a change of vendor.
Terms
- Digital twin
- A virtual model of a physical asset or system, updated from live data and used to support decisions about it.
- SCADA
- Supervisory control and data acquisition: the systems utilities and plants use to monitor and control equipment.
- Phasor measurement unit (PMU)
- A grid sensor that measures voltage and current many times a second, time-stamped by satellite.
- BIM
- Building information modelling: a detailed digital model of a building's design, often the starting point for a building twin.
- VVUQ
- Verification, validation and uncertainty quantification - checking a model is built right, matches reality, and says how sure it is.
Steepens
| Megatrend | Weight | Role |
|---|---|---|
| Hyper-automation of Industrial Production | moderate | shared weakest link (with 3 others) |
| Energy Transition and Electrification | moderate | |
| The Built-Environment Deficit | moderate | shared weakest link (with 7 others) |
| Precision Everything | moderate | shared weakest link (with 4 others) |
The gradient behaves like a sticky rubber band: it is pulled hardest toward the weakest enabler. Marginal links are shown but do not pull.
If this moved one step
Computed from the map as it stands: this score raised from 2 to 3, every other score unchanged. It is arithmetic, not a forecast.
Raised one step on its own, it lifts no megatrend's weakest link; each gradient it feeds still rises a little.
Tied as the weakest link
Here it shares the weakest spot with other enablers. Raising it alone leaves the weakest link where it is and lifts the gradient a little; the enablers it is tied with must move too.
- Hyper-automation of Industrial Production2.8/5 , 2.9/5tied with Digital Transformation, Managerial Capability, Workforce Trust
- Precision Everything2.7/5 , 2.8/5tied with Data & Datafication, Managerial Capability, Smart Cities, Workforce Trust
- The Built-Environment Deficit2.4/5 , 2.5/5tied with Biomaterials & Biofabrication, Construction Technology, Family Formation, Land Use & Housing Supply, Permitting Speed, Public Procurement, Smart Cities
The gradient also rises in
- Energy Transition and Electrification2.3/5 , 2.3/5
What would change the gradient
Conditional, not predictive. No dates, no probabilities.
- 01Sensor density on the underlying asset, which is the real cost
- 02Model maintenance as an ongoing obligation rather than a project deliverable
- 03Standards, so a twin survives a change of vendor
Common misconception
That a twin is a model. A model without a live data feed is a simulation; the twin's value is the synchronisation, and the synchronisation is the expensive part.
Talin, B. (2026). Digital Twins. The Future Explorer, dataset v0.4.0 (2026-11-16). https://explorer.talin.digital/e/digital-twins/