energy · infrastructure · family: the model was never trained on this
the building has no twin
Most buildings that need retrofits have no digital models to optimize them
Problem statement
Digital twins — real-time virtual replicas of physical buildings — can optimize retrofit decisions, predict energy savings, and adaptively control HVAC systems; several studies report energy reductions of more than 20–30% where they are deployed, while other studies find only marginal gains once commissioning quality and occupant behaviour are taken into account. But the buildings most urgently needing energy retrofits are precisely the ones that cannot use digital twins: older structures that lack Building Information Models (BIM), sensor networks, or any form of digital representation. The EU's building renovation rate stands at roughly 1% per year, meaning at current pace it would take a century to retrofit the existing building stock. Digital twin technology could accelerate this rate by making retrofit decisions faster, cheaper, and more accurate — but it requires a digital foundation that most buildings simply don't have.
Why this matters
Buildings and construction consumed about 32% of global energy and produced about 34% of global energy-related CO₂ emissions in 2023; the digital twin review puts the building sector at nearly 40% of energy-related greenhouse gas emissions. The IEA reports that despite COP28 pledges, global energy efficiency improvement was only ~1% in 2024, and that the world is not yet on track for the COP28 goal of doubling the global average annual rate by 2030; the primary review says the pace needs to accelerate to about 4% per year to meet climate goals. The IEA also reports that almost four out of five construction companies globally say they experience a shortage of skilled workers. The retrofit gap is not only a technology problem — the primary review concludes that most of the technologies already exist — but also a decision-support problem: building owners and municipalities cannot cost-effectively determine which retrofits will deliver the best returns for which buildings, especially across large heterogeneous building portfolios.
What’s been tried and why it hasn’t worked
Most digital twin research and deployment focuses on new construction, where BIM models are created during design. For existing buildings, creating a digital twin requires installing sensors and meters and calibrating a model — the sources describe this integration as expensive and complicated, name cost as a barrier for small and older buildings, and warn that extending digital twins to large portfolios or heterogeneous building stocks may force partial rather than full-fidelity twins. Static energy audits, rule-based control and periodic retrofitting are the conventional approach; the digital twin review says they have delivered measurable gains but are limited in scope, assume fixed operating conditions, and fail to capture how energy performance shifts with occupancy, weather, internal loads and equipment degradation. The DanRETwin project in Denmark proposed a digital twin solution for existing non-residential buildings that combines dynamic building energy models with real-time data from energy meters and clamp-on IoT sensors — an alternative, temporary means of collecting data from buildings that have no sensing or metering of their own.
What would unlock progress
A breakthrough likely requires automated, low-cost methods for generating "lightweight" digital twins of existing buildings — models that capture enough building physics for useful retrofit optimization without requiring full BIM detail. This could combine exterior scanning (widely available via aerial imagery and LiDAR), utility billing data (available for most buildings), and occupant feedback into simplified thermal models that improve over time through machine learning. The approach would need to work with heterogeneous building stocks across different eras, construction types, and climate zones. Middleware platforms using open standards (the digital twin review names NGSI-LD, IFC and SAREF) could bridge legacy building management systems with modern digital twin frameworks.
Entry points for student teams
A student team could prototype a "minimal viable digital twin" for an existing campus building using only publicly available data: building footprint from GIS, estimated age and construction type from tax records, weather data from nearby stations, and monthly utility bills. The team would build a simplified thermal model, calibrate it against actual energy use, then simulate the impact of specific retrofit measures (window upgrades, insulation, HVAC replacement). Comparing the model's predictions against actual post-retrofit performance data (where the campus facilities office can share it) would quantify the accuracy tradeoff. Skills in building science, data science, and software engineering would be most relevant.
Genes
Structural cousins: stuck the same way, in other fields
Sources
"Energy Efficiency and Decarbonization Strategies in Buildings: A Review of Technologies, Policies, and Future Directions," *Applied Sciences*, MDPI, 15(21):11660, 2025. (accessed 2026-02-11). Supplemented with "Unlocking the Potential of Digital Twin Technology for Energy-Efficient and Sustainable Buildings: Challenges, Opportunities, and Pathways to Adoption," *Sustainability*, MDPI, 18(1):541, 2026; and "DanRETwin: A Digital Twin Solution for Optimal Energy Retrofit Decision-Making and Decarbonization of the Danish Building Stock," *Applied Sciences*, MDPI, 13(17):9778, 2023. go to source ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
- The IEA Energy Efficiency 2024 report quantifies the gap between pledges and actual progress: https://www.iea.org/reports/energy-efficiency-2024/executive-summary
- The digital twin review's Table 4 gives estimated emissions-reduction potentials by use case: 10–30% for predictive HVAC control, 5–15% for fault detection and diagnostics, 10–20% for demand-side management, 20–40% for retrofit scenario optimization and 15–25% for renewable energy integration. The review says these figures reflect optimal or model-based scenarios, and that actual savings depend on the building data available, the extent of deployment and the context.
- Cross-domain connection: this problem shares structure with the infrastructure-cascading-failure-modeling brief, where digital representations of physical systems are needed for predictive analysis but don't exist at the required scale.
- The workforce shortage (almost four out of five construction companies reporting a shortage of skilled workers, per the IEA) compounds the problem: even manual energy audits are bottlenecked by available talent.
- Related to the "building performance prediction gap" problem — even when digital models exist, their accuracy is limited by static assumptions.
Reconciliation 2026-09-30: Fact repair after reading the whole write-up against every source it names. The three MDPI journal pages return 403 to scripts, so all three papers were read in full on 2026-09-30 from the publisher's PDF mirror: the primary review (Jørgensen and Ma, Applied Sciences 15(21):11660, 56 pages, https://mdpi-res.com/d_attachment/applsci/applsci-15-11660/article_deploy/applsci-15-11660.pdf), the digital twin review (Jradi, Sustainability 18(1):541, 28 pages, https://mdpi-res.com/d_attachment/sustainability/sustainability-18-00541/article_deploy/sustainability-18-00541.pdf) and the DanRETwin paper (Jradi, Madsen and Kaiser, Applied Sciences 13(17):9778, 20 pages, https://mdpi-res.com/d_attachment/applsci/applsci-13-09778/article_deploy/applsci-13-09778.pdf); their citation records were fetched from Crossref by DOI. The IEA Energy Efficiency 2024 executive summary was read in full directly from iea.org. (1) The text said digital twins deliver "30–40% energy efficiency improvements in buildings where they're deployed." No source gives that range for digital twins; the digital twin review says several studies report reductions exceeding 20–30% through predictive control and optimization while other empirical studies find marginal gains once commissioning quality and occupant behaviour are considered. The sentence now says that. (2) The text said buildings account for "approximately 36% of global energy consumption and nearly 40% of energy-related CO₂ emissions." The primary review gives about 32% of global energy and about 34% of global energy-related CO₂ for 2023; the digital twin review gives nearly 40% of energy-related greenhouse gas emissions; the 36% figure appears only in the DanRETwin paper, as the EU share of CO₂, not global energy. The sentence now gives the two sources' figures. (3) The IEA page does not state a "4% annual target"; it describes the COP28 goal of doubling the global rate of efficiency improvement by 2030, and says the world is not yet on track to achieve it; the primary review separately says the pace of improvement needs to accelerate to about 4% per year to meet climate goals, without tying that figure to COP28. The sentence now gives the two statements side by side. (4) The text said the IEA found "72% of employers in energy efficiency roles report skilled workforce shortages." No source contains 72%; the IEA page says almost four out of five construction companies globally indicate they experience a shortage in skilled workers. The sentence and the Source Notes bullet now use the IEA's wording. (5) "Not primarily a technology problem … but a decision-support problem" was stronger than the sources; the primary review concludes most technologies exist and are under-used because of fragmented policies, limited retrofit rates and behavioural barriers. The sentence now says "not only … but also." (6) The cost sentence said creating a twin requires "manual surveying, 3D scanning, sensor installation, and model calibration — a process costing tens of thousands of dollars per building." No source gives a dollar figure or mentions surveying or scanning. The DanRETwin paper says sensing and metering devices are generally not available in existing buildings and their integration is expensive and complicated; the digital twin review lists high initial costs as a barrier for small and older buildings and says scaling to large portfolios or heterogeneous stocks may necessitate partial rather than full-fidelity twins. The sentence now says that. (7) The sentence about LiDAR and photogrammetry capturing "only the envelope, not the internal systems" is in none of the sources and was removed. (8) The text said rule-based energy audits are "labor-intensive, provide only static snapshots, and frequently miss interactions between building systems." The digital twin review says static energy audits, rule-based control and periodic retrofitting have delivered measurable gains but are limited in scope, assume fixed operating conditions, and fail to capture how performance fluctuates with occupancy, weather, internal loads and equipment degradation; "labor-intensive" and "miss interactions" are not there. The sentence now follows the review. (9) The text said DanRETwin "demonstrated" a twin for the Danish stock but "relied on standardized building archetypes rather than actual building data." The paper is a design proposal (it says the solution "will" provide its services) and describes the opposite data approach: real-time data from energy meters and clamp-on IoT sensors combined with dynamic energy models, offered as an alternative, temporary means of collecting data from buildings that lack sensing. The sentence now says that. (10) "OPC UA, MQTT" appear in no source; the digital twin review's open standards are NGSI-LD, IFC and SAREF, and the parenthesis now names those. (11) The entry point said post-retrofit performance data is "available for many university buildings"; no source says so, and the parenthesis now says "where the campus facilities office can share it." (12) The Source Notes bullet giving "15–25% carbon emission reduction in industrial settings and 30–40% improvement in building energy efficiency" matched nothing in the sources; it now reports the digital twin review's Table 4 potentials by use case (10–30% predictive HVAC control, 5–15% fault detection, 10–20% demand-side management, 20–40% retrofit scenario optimization, 15–25% renewable energy integration), with the review's caveat that these reflect optimal or model-based scenarios and that actual savings depend on building data, the extent of deployment and context. (13) The Source line gave the second and third papers' titles without their subtitles; the full titles were copied from the Crossref records. Confirmed in the sources and left as written: the primary citation (Applied Sciences 15(21):11660, 2025) and the other two volumes, issues, article numbers and years; that digital twin work focuses on newly constructed buildings with BIM while existing and older buildings often lack any form of digital modelling (DanRETwin paper); the EU renovation rate of around 1% of the stock per year (DanRETwin paper) and the primary review's statement that at around 1% per year it would take a century to upgrade the stock; the IEA's figure of about 1% global efficiency improvement in 2024, the same as 2023; the digital twin review's statements on middleware platforms, legacy building management systems and heterogeneous stocks; and the primary review's description of calibrating a model to utility data for retrofit analysis. The 1% renovation figure is given for the EU in the DanRETwin paper and for the world in the primary review, and the write-up's "a century" is the primary review's phrase; this is left as written. Not checked, because they are design proposals rather than source claims: the "lightweight" twin combining aerial imagery, utility bills and occupant feedback, and the student entry point. No title changed. Genome tags untouched. A second reader went over these repairs against the same four sources on the same day; items (3) and (12) are given here in the second reader's final wording, and the rest stand as first repaired.