Ensuring trust with explainable AI for building energy management.
BlockDox, UCL, Portakabin, Verv, D-Fine
The project is supported by Innovate UK under the Accelerating trustworthy AI: Phase 2 Collaborative R&D competition – project number 10093095 (March 2024 – Feb 2025)
Duncan Wilson, Steve Gray
TRUST Phase 2 aims to advance understanding of the complex interactions between physical spaces, technical systems and human occupants in modular buildings, using commercially proven sensors and equipment to develop explainable AI/ML models for optimized building management.
The goal of the project is to provide the tools to support:
Artificial Intelligence (AI) and Machine Learning (ML) hold the promise of revolutionizing the built environment, from construction and retrofitting to ongoing operations. They offer the potential for substantial cost savings, energy optimization, and even revenue generation. Yet, despite these compelling benefits, the adoption of AI and ML in buildings has been surprisingly slow and often limited to isolated applications. One exception is the use of AI for equipment fault detection and computer vision sensors with AI edge processing, which have seen more widespread implementation. However, technologies that could significantly reduce a building’s carbon footprint – such as those that measure and predict occupancy to optimize heating and cooling systems – remain underutilized. The concept of a truly “Smart Building” – where integrated AI and ML solutions optimize building controls in real time – remains largely aspirational. While a few such buildings exist, they are the exception, not the rule.
Whilst this research can be applied across many different building types, this project will focus on developing a best-in-class turnkey solution for mass-market adoption in modular buildings. Modular buildings are prefabricated structures constructed off-site and then assembled at their final location. They offer a versatile and efficient alternative to traditional construction methods, with various applications across different sectors such as construction (site offices), health care (clinics) and education (classrooms). These buildings have the common core elements of permanent building structures (power, light, heating/cooling, water), and are portable only in terms of location, being installed for weeks, months or years at each site during their 30+ year lifespan. Energy consumption is often overlooked in these buildings, with equipment, heating or cooling left running unnecessarily or at incorrect set points. Remote and/or automated control of heating and cooling systems can save costs and reduce the carbon footprint of their portable buildings.
TRUST Phase 2 is a collaboration between two tech SMEs (BlockDox and Verv), each with existing AI/ML-based products, to develop a system that incorporates and advances their respective AI/ML technologies. This system will be deployed in several modular buildings across the UK as operational pilots in real-world environments. Portakabin, a leading modular building manufacturer with a large rental business, will provide the test and demonstration units.Operational parameters such as heating and cooling will be controlled in response to varying occupancy levels and indoor and outdoor air quality conditions to achieve optimized demand-side management. The project will measure the ROI and benefits, particularly regarding energy efficiency and operational cost savings.
BlockDox and UCL will collaborate to develop simulation models of the portable buildings using advanced AI/ML modeling techniques. These models will facilitate the development of prediction algorithms and enable testing of control strategies in silico, providing a wide range of test conditions not achievable in real-world scenarios. The models will also provide confidence to end users that the system will perform as expected, even under atypical operating conditions.
D-Fine will provide oversight and assurance that the AI/ML algorithms are explainable, trustworthy, and responsible. The AI assurance partner will assess the system for explainability, transparency, trustworthiness, and responsibility. The pilot’s output will be a series of compelling case studies as the project deliverables.
