For projects within the UK, CIBSE TM54 and the NABERS UK DfP framework present practical and effective tools that enable design teams to progress beyond simple regulatory compliance towards buildings that genuinely achieve their intended in-use energy performance.
Both approaches advocate a performance-led design philosophy by proactively addressing uncertainties, accurately refining assumptions, and integrating realistic system modelling throughout the building’s life-cycle.
CIBSE TM54 offers a structured methodology specifically designed for the evaluation of operational energy use during a building project’s design stages. The primary aims are to facilitate the creation of buildings aligned with ambitious net zero carbon targets, and to significantly reduce the well-documented discrepancy between predicted and actual energy consumption – the EPG.
This is achieved by encouraging modellers to consider the inherent uncertainty in input parameters and the complex behaviour of building systems – representing an important shift from typical ‘design-for-compliance’ approach. These are often based on simplified methodologies such as required by Part L in England and Wales, Scotland’s Section 6, or Northern Ireland’s booklets F1/F3, and tools like the standard assessment procedure (SAP) or simplified building energy model (SBEM).
The TM54 methodology provides flexibility by accommodating a range of modelling complexities, from spreadsheet-based estimations to steady-state methods and dynamic simulation modelling (DSM). This allows for the selection of the most appropriate level of detail based on the project’s specific complexity and available resources.
A key strength of TM54 is its comprehensive scope, encompassing both regulated energy loads (such as space heating, hot water, cooling, fans, pumps and lighting) and typically unregulated loads (including office equipment, IT servers, lifts and catering). For mixed-use or multi-tenant buildings, TM54 highlights the need for clear and precise differentiation between base building (landlord-controlled) and tenant-specific energy consumption.
A key characteristic of TM54 is its integration across the entire building project life-cycle. Commencing at the initial briefing stage, the methodology evolves through concept design, detailed design, construction, commissioning, handover and, ultimately, into in-use operation. The design-stage TM54 model is intended as a living document, iteratively updated as the project progresses, culminating in an as-built energy model.
This provides a baseline for post-occupancy evaluation (POE), particularly in conjunction with CIBSE TM633, which offers specific guidance on evaluating energy performance in occupied buildings. Consequently, TM54 actively supports continuous, data-driven commissioning and ongoing fine-tuning of building systems based on this evolving model.
The TM54 methodology is structured around 17 core steps, as summarised in Table 1.
The NABERS UK DfP framework closely aligns with the principles of TM54, and is specifically targeted at office buildings. A significant step forward is the contractual, pre-construction commitment made by the developer or building owner to achieve a specific NABERS UK energy base building rating once the building becomes operational. This – hopefully ambitious – target then serves as a central and unwavering reference point throughout the entire design, construction and operational phases of the project.
In contrast to compliance-driven modelling, NABERS UK DfP mandates the use of advanced dynamic simulation to estimate base building energy consumption with a high degree of accuracy. These simulations must account for hourly operational profiles, internal heat gains, the thermal properties of the building fabric, thermal inertia and, crucially, the detailed dynamic behaviour of heating, ventilation and air conditioning (HVAC) systems and their associated controls under realistic operating conditions.
This necessitates modelling that incorporates part-load performance characteristics, seasonal variations in efficiency and the inherent losses associated with real-world operation.
To ensure the robustness and credibility of the DfP process, NABERS UK requires an Independent Design Review (IDR) conducted by qualified and accredited reviewers. These independent experts critically assess the modelling approach, the detailed representation of HVAC systems, the proposed energy metering strategy, and the overall likelihood of the building achieving the agreed-upon energy target.
The IDR plays a vital role in ensuring consistency with DfP principles, and in identifying potential risks and opportunities for further energy optimisation.
Echoing TM54’s emphasis on presenting a range of potential outcomes, scenario testing is a fundamental element of NABERS DfP. The framework requires the model to explore ‘off-axis’ scenarios that represent potential variations in control logic, unforeseen specification gaps, delays in the commissioning process, deviations in occupancy patterns, or unexpected operational strategies. DfP also mandates the inclusion of a modelling margin to account for any residual uncertainties that may still exist.
The developed building systems will almost certainly include – possibly intelligent – controls that adjust output to match dynamic demand to improve part-load efficiency. However, standard DSM templates may not capture these nuances. The increasing availability of building technologies, including building management systems (BMS) with Internet of Things (IoT) sensors, and digital twins, offers the potential for even deeper insights into energy consumption patterns and occupant behaviour.
When these technologies are combined with the power of artificial intelligence (AI), machine learning (ML) algorithms, and ‘sensor fusion’ (combining data from multiple sensors) techniques, they can enhance the quality and granularity of design inputs. This enables the development of more sophisticated models that more accurately reflect how buildings are occupied and operated.
Tools such as building information modelling (BIM) and virtual reality/augmented reality (VR/AR) are increasingly popular for engaging stakeholder engagement, as well as informing iterative design processes.
Shared data platforms, the adoption of open data standards and the development of consistent ontologies can facilitate more effective coordination and the reuse of high-quality occupancy and performance data. The implementation of IDRs provides an additional critical layer of scrutiny, helping to verify the rigour of the modelling process, the realism of underlying assumptions and the suitability of the proposed energy metering – and submetering – strategies.
Placing the needs and behaviours of occupants at the very centre of energy strategies is fundamental. The increased application of occupant-centric controls (OCC) – informed by behavioural research, intuitive user interfaces, and real-time feedback mechanisms – can lead to building systems that dynamically adapt to actual occupant needs, ultimately improving both comfort and energy efficiency.4
The development of standards and guidance for accurately modelling human-building interactions, alongside updates to building codes that better reflect occupant behaviour and internal environmental quality (IEQ) considerations, will support more consistent and effective implementation of performance-based design.
Despite the significant progress represented by TM54 and NABERS DfP, several persistent limitations in current modelling practices and process integration continue to contribute to the EPG. Traditional simulation tools often lack the inherent sophistication required to accurately represent the intricate control strategies of modern HVAC systems, such as the heat pump in Figure 1, their part-load efficiencies across a range of operating conditions, and the complexities of real-world operational dynamics.
Figure 1: Monobloc air source heat pump (40–140 kW) using natural R-290 refrigerant, providing heating up to 75°C and operation down to -25°C outdoor temperature, with seasonal coefficient of performance (SCOP) up to 4.20 and energy efficiency up to 30% above Ecodesign requirements (Source: Carrier)
A reliance on simplified or default assumptions regarding occupancy patterns, internal heat gains from equipment, and building management practices can lead to unreliable energy projections. The very real constraints of time and budget in practical projects can also limit the extent to which highly detailed modelling can be realistically undertaken.
A further challenge lies in the absence of a widely adopted and standardised data model for representing the multifaceted aspects of occupant-related behaviour within building performance simulations. This deficiency hinders the effective integration of both qualitative data insights (from occupant surveys) and quantitative data (from building sensors) into building performance models.
Consequently, the significant influence of occupant behaviour – a critical driver of actual energy consumption – is often underestimated or inaccurately represented in design-stage predictions.
Process-related issues, including fragmented communication pathways and a lack of clear accountability across the various stages of design, construction, and ongoing operation, further exacerbate the EPG. The absence of consistent protocols for systematically capturing and modelling occupant behaviour compounds this challenge.
The increasing prevalence of smart, connected HVAC systems presents valuable opportunities for continuous performance monitoring, automated fault detection and ongoing energy performance optimisation. These advanced capabilities enable building operators to directly compare in-use energy data with the original design-stage predictions, providing crucial feedback that supports ongoing commissioning activities and helps to identify any operational inefficiencies or control-related issues.
Both TM54 and TM63 explicitly promote the concept of design-stage energy models evolving into valuable operational tools. With appropriate documentation and calibration using real-world performance data, the original simulation model can be effectively utilised during post-occupancy evaluation to fine-tune control strategies, provide robust data for energy audits and inform strategic retrofit decisions.
This integrated approach fosters a more feedback-driven process, closely aligning with the objective of achieving performance-in-use targets and providing a clear pathway towards net zero operational outcomes.
The ability to understand and accurately predict operational energy use at the design stage is continually being enhanced by several evolving techniques and ongoing research directions, primarily focused on improving the quality of input data, the sophistication of model architectures and the overall accuracy of key assumptions.
ML techniques are playing an increasingly significant role in building design, offering powerful capabilities for predicting energy performance, optimising system control strategies and supporting faster – more informed – decision-making. Among these, artificial neural networks (ANNs) have demonstrated the ability to learn from large datasets of building simulations to accurately estimate energy consumption for new designs.
ANNs are already being applied in the early stages of architectural design to inform building form, identify key factors influencing energy consumption, select appropriate structural systems and develop realistic energy cost budgets.5 While feedforward neural networks (FFNNs) have historically been the dominant architecture in this area, convolutional neural networks (CNNs) – designed to detect spatial patterns in data – have gained significant traction in recent years.
In building energy modelling, CNNs are increasingly being used to capture complex spatial relationships between various design features (such as building layout or orientation) and overall energy performance. Comparative analyses at the design stage have shown that ANN-based models often achieve higher accuracy and efficiency compared with other ML methods, such as support vector regression and long short-term memory networks.4
Emerging data-driven techniques are unlocking new possibilities for improving the accuracy of operational energy predictions during building design. Advancements in data processing – including sophisticated techniques for data cleaning (identifying and rectifying inconsistencies and errors), feature selection (identifying the most relevant input variables), and data integration (combining data from disparate sources) – are significantly improving the quality of inputs used in simulation models.
Furthermore, the exploration of real-world data harvested from smart buildings and the application of generative models such as generative adversarial networks (GANs)6 are enriching design-stage datasets. GANs – which learn to generate new, realistic data by pitting two neural networks against each other – can be particularly valuable in supplementing limited real-world datasets, thereby improving the training of predictive models even when actual performance data is scarce.
More advanced ML architectures are also gaining attention, including hybrid models that combine the strengths of different approaches, and generative pre-trained transformers (GPTs).5 Originally developed for natural language processing, GPTs excel7 at analysing sequential data, such as extracting valuable insights from post-occupancy surveys, interpreting building regulations or processing design documents to support automation and enhance decision-making.8,9,10 GPTs have the potential to uncover hidden patterns within large datasets, leading to more robust and reliable energy forecasting.
Ongoing research is also heavily focused on optimising the parameters of these complex models, refining critical input assumptions related to occupancy patterns and climate data, and improving the interpretability of the models themselves – ensuring that even sophisticated AI-driven predictions remain transparent and useful for informing design decisions. Collectively, these innovations are playing a crucial role in bridging the EPG and supporting a more accurate and truly performance-led approach to building design.
Together, CIBSE TM54 and NABERS UK DfP provide a comprehensive and robust framework for evaluating, and ultimately achieving, meaningful improvements in operational energy performance. By emphasising the integration of rigorous modelling techniques, comprehensive scenario analysis, the use of realistic building services plant data, and the crucial role of in-use performance monitoring, these methodologies are instrumental in helping to ensure that buildings not only meet their initial design intent but also perform as intended in real-world operation.
© Tim Dwyer 2025.
References:
1 CIBSE TM54 Evaluating operational energy use at the design stage, 2nd ed CIBSE 2022.
2 NABERS UK Guide to Design for Performance, Version 3.0, CIBSE 2024 – bit.ly/CJJun25CPD1
3 CIBSE TM63 Operational performance – Evaluating operational energy use, CIBSE 2022.
4 Wagner, A and O’Brien, L, Occupant-Centric Building Design and Operation, Annex 79 Final Report 2024 – bit.ly/CJJun25CPD2
5 Yin, Q et al, A Review of Research on Building Energy Consumption Prediction Models Based on Artificial Neural Networks, Sustainability 2024, 16, 7805 – bit.ly/CJJun25CPD3
6 Chen, G et al, A Systematic Review of Building Energy Consumption Prediction: From Perspectives of Load Classification, Data-Driven Frameworks, and Future Directions, Appl. Sci. 2025, 15, 3086 – bit.ly/CJJun25CPD4
7 Brown, TB et al, Language models are few-shot learners, Adv Neural Inf Process Syst. 2020 33:1877–1901
8 Liu, M et al, Large language models for building energy applications: Opportunities and challenges, Build Simul 2025 18: 225–234 – bit.ly/CJJun25CPD5
9 Chen, N et al, Automated Building Information Modeling Compliance Check through a Large Language Model Combined with Deep Learning and Ontology, Buildings 2024, 14, 1983 – bit.ly/CJJun25CPD6
10 Fuchs, S et al, Using Large Language Models for the Interpretation of Building Regulations, 13th Conference on Engineering, Project and Production Management, 2023, Auckland, NZ – bit.ly/CJJun25CPD7
