Digital twins, combined with edge AI, have proven capable of reducing the operational expenditure of smart buildings.
In commercial real estate and campus environments, “phantom load” (energy consumed by devices in standby or idle mode) can account for as much as 32 percent of a building’s total energy profile. Previous research indicates that up to a third of electricity used in office buildings is attributable to this phantom power.
For business leaders, the first step is often an audit of these “always-on” assets to identify immediate savings opportunities. While many enterprises have adopted high-level metering, specific control at the plug level remains difficult due to the high coordination costs of decentralised device management.
Engineers from the University of Glasgow’s James Watt School of Engineering have developed a prototype of a digital tool designed to tackle this waste without disrupting business continuity.
The impact of phantom load
The concept of phantom load is often dismissed as negligible, yet the aggregate impact on a corporate P&L is substantial. Research indicates that plug-in devices – ranging from monitors and workstations to servers – comprise a major portion of building energy use. In student housing alone, standby power can represent up to 33 percent of total electricity usage.
Dr Ahmad Taha, Lecturer for Autonomous Systems & Connectivity at the James Watt School of Engineering, who is leading the work, said: “I’m a firm believer in the idea that that small, collective actions on climate issues can have big effects, and phantom power use is an obvious candidate for that kind of action.”
The difficulty often lies in distinguishing between a device that is idly wasting power and one that is in a necessary low-power state for rapid reactivation. Traditional binary control systems (timer-based on/off) often fail because they lack context, leading to user frustration and eventual overriding of the system.
Adoption of these controls increases when logic accounts for user habits and probability of return, rather than relying on simple schedule-based switches. The proposed Edge-Enabled Digital Twins (EEDT) system for smart buildings addresses this by creating a virtual representation of physical assets on a local edge server where AI can be used for additional insights and automation.
By processing data locally rather than in the cloud, EEDT also lowers privacy risks associated with monitoring individual usage patterns while ensuring the low latency required for real-time control. Prioritising this local edge processing is essential to resolving employee privacy concerns while unlocking the potential of AI.
The core differentiator in this approach is the move away from rule-based automation toward “fuzzy logic” (a computing approach based on degrees of truth rather than the usual true or false Boolean logic.) The system draws data from a network of smart energy sensors, which send information on electricity to a central server using the LoRaWAN protocol that is widely-used for IoT systems.
The prototype utilises a decision-making framework based on 27 optimised rules. Instead of simply cutting power after a set time, the system calculates three specific metrics:
- User Habit Score: This analyses usage likelihood and stability to understand behavioural routines.
- Device Activity Score: This integrates standby duration and time since the last active state to assess current inactivity.
- Confidence Score: This gauges data reliability to ensure the system does not act on incomplete information.
These inputs allow the digital twin to make flexible decisions about the assets of smart buildings: immediate shutdown, delayed decision, user notification, or maintaining the current state. When the system detects prolonged idle periods, it sends users a prompt on their screen to determine if they’re conducting remote work or running background processes.
This approach aims to raise user’s awareness of their device’s idle periods, perhaps encouraging them to make more careful use of their devices, while also preventing legitimate work processes from being cut off.
Operational results and ROI
To validate the architecture, the researchers deployed the system in a university research laboratory, utilising smart plugs and environmental sensors communicating via LoRaWAN.
The results offer a solid business case for intelligent edge AI-powered management using digital twins. The deployment demonstrated a reduction in weekly power consumption of approximately 40.14 percent per monitored workstation. Specifically targeting phantom loads, the fuzzy decision-making framework achieved a reduction of up to 82 percent.
When extrapolated to a wider smart buildings deployment, the financial implications become evident. Based on the UK electricity price cap as of July 2025, deploying this system across 500 devices is projected to yield annual savings exceeding £9,000.
Beyond immediate energy savings, Dr Taha highlights a secondary financial benefit regarding asset lifecycle management: “Secondly, by reducing devices’ use of electricity, it could help reduce the need to replace older devices with newer, more power-efficient ones.
“That in turn could help organisations save on equipment costs in an increasingly challenging economic environment.”
The technical implementation of such a system often relies on a containerised edge architecture. The research team utilised Docker containers hosting an MQTT broker for messaging, Node-RED for data parsing, and InfluxDB for time-series storage. This stack allows for “closed-loop” control, where the digital twin not only monitors but actively intervenes in the physical world.
A necessary component for user acceptance is the ‘Anti-Oscillation Filter’. In early automated systems, rapid switching between on and off states (hysteresis) often caused hardware wear and user annoyance. The EEDT system incorporates cooldown management and stability checks to ensure that a decision to shut down a device is stable and contextually appropriate.
The system also integrates a forecasting module using Long Short-Term Memory (LSTM) deep learning. By training on just two days of historical data, the model predicts the next day’s consumption trend. Integrating these short-term prediction models allows facilities teams to anticipate peak loads rather than just reacting to them.
Edge AI-powered digital twins: Making buildings truly smart?
The transition from passive energy monitoring to edge AI-driven optimisation using digital twins is the next necessary step for smart buildings. While this study focuses on a university setting, the architecture is directly transferable to corporate offices, healthcare facilities, and industrial environments where plug loads remain unmanaged.
Dr Taha added: “Reaching net-zero will require a broad-spectrum approach to energy monitoring, and this tool could be a valuable part of wider institutional approaches to minimising their carbon footprint using digital twins to monitor variables like occupancy and temperature control.” The team is currently working to investigate how this tool could play a role in the University’s wider efforts to achieve net-zero by 2030.
Scaling requires addressing legacy infrastructure. The reliance on manually designed fuzzy rules (27 in this specific case) may limit rapid scalability across diverse asset types. Future enterprise-grade solutions will likely need to incorporate neuro-fuzzy learning to automate rule generation based on specific departmental behaviours.
The data required to cut energy costs exists within the network. The challenge is no longer about gathering that data and visualising it with digital twins, but about empowering edge assets with AI to act on it intelligently and make buildings truly smart.
See also: Industrial AIoT adoption drives operational efficiency
Want to learn more about IoT from industry leaders? Check out IoT Tech Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including AI & Big Data Expo and the Cyber Security Expo. Click here for more information.
IoT News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
