We cover how the Home Energy Model has been validated against other models and real buildings.
In this article, we cover how the Home Energy Model (HEM) has been validated.
Validation ensures HEM meets its goals. Unlike models used in domestic settings today (PHPP and SAP), HEM can model half hourly building energy use. Validation ensures this capability can be trusted to precisely model renewable systems like heat pumps and solar PV, detect peak events such as overheating or cold snaps, and reflect nuanced occupant behaviour and heating control use.
The main way HEM has been validated is by comparison to other energy models. HEM outputs have also been compared to the measured performance of several real life buildings, and in lab trials. These comparisons are also covered in brief here, and further reading can be found at the end of this article.
Comparing HEM to other models enables a systematic understanding of its differences, in a manner that eliminates real-life noise and isolates model characteristics. For this reason, a “shoebox” (simple rectangular prism) building was modelled before the more complex building shapes to isolate differences caused by core building physics, and aligned assumptions were tested before each model’s “conventions'' were applied to understand the effect of these on output.
HEM's half hourly outputs limits the ability for PHPP (which has monthly outputs) to be a benchmark for model predictions. HEM was therefore primarily compared to ESP-r, a dynamic model known for robust modelling of building physics. Aggregated monthly and annual figures were also compared to Passivhaus Planning Package (PHPP), a static model known for its reliably predicting real world performance.
While dynamic models are the gold standard of simulating building physics, static models have clear advantages - they can operate with limited information and lower computation needs, making them useful field tools capable of validation against larger real-life datasets.
Previous static models have been low granularity, making them poor tools at understanding the complex interaction between a building's fabric, systems, occupants and environment. HEM’s comparability to dynamic models makes it uniquely capable amongst static models, capable of bridging the gap between building physics and real life performance.
The comparison between HEM and other models occurred across two phases. Phase 1 involved using the same design and aligned inputs for each model, so that differences in outputs could only be driven by methodology.
The key points noted at this stage:
In Phase 2, models were compared but with standard conventions applied. In HEM’s case, this meant applying the assumptions of the Future Homes Standard wrapper. This caused greater divergence between models - as expected, given different assumptions used for important variables such as unregulated energy use or ventilation.
The HEM team succinctly summarised the outcome of the two phases: “Phase 1 of this IMC study has indicated that the HEM core methodology is suitably aligned with the comparator models; however, under its intended Future Homes Standard application, as was the case in Phase 2, the level of agreement was much reduced.“
This conclusion on the impact of Future Homes Standard assumptions on HEM was also reflected by LETI’s analysis, noted below.
The official HEM validation studies include several real-world tests - two Fraunhofer Institute buildings in Holzkirchen Germany (as part of an IEA funded project), three buildings in Marmalade Lane in Cambridge (as part of the Building for 2050 project), and the Camden Passivhaus. These homes were selected due to the rich information available on each building - the Fraunhofer buildings were also unoccupied, making them even easier to precisely model.
These studies leveraged the Home Energy Model’s flexibility - instead of using Future Homes Standards assumptions for unregulated energy use or weather patterns, real world values were used for weather, occupancy and space heating regime.
Once calibrated, HEM gave very similar results to reality for all of these test homes, minus areas where the model was in development (eg, the Camden Passivhaus solar thermal system could not be modelled at the time). For example, Part L regulated energy use was brought within 8% of measured data, with the biggest discrepancies coming from heating system heat losses and auxiliary HVAC equipment operation.
The alignment was best for annual data - with the greater discrepancy in monthly and daily data, which the HEM team speculated could be caused by differences between modelled data and reality. For example, in the Camden Passivhaus study the team noted that it was not clear from measured data how exactly occupants set heating controls, how the use of blinds impacted solar gains, or even exactly when occupants were in the building versus on holiday. This made real world trials interesting case studies, but limited their usefulness for model validation.
LETI published a FHS consultation response where they correctly noted that the Future Homes Standard was not close to representing a best-in-class energy efficient home. In their analysis, they noted that the notional buildings provided by HEM had 2.5 - 5x the energy use of genuine low energy buildings, and that the FHS assumptions understated the space heating demand of these buildings versus PHPP.
The two inaccuracies LETI points out for the Home Energy Model (ventilation and unregulated energy use) are associated with Future Homes Standard assumptions, NOT the core HEM modelling engine. As noted in the inter-modal validation and the real world trials, HEM has flexible assumptions including for ventilation and unregulated energy use.
Unlike previous models used in domestic settings HEM can produce comparable results to dynamic energy models - though there are gaps to dynamic methodologies driven by a lack of dynamic HTC and thermal mass modelling. Nevertheless, it is capable at identifying peak solar gain and shading events, in a way not possible with SAP or PHPP. This is important to capture overheating events, or understand generation or heating system performance.
HEM has also been compared to real life buildings. This demonstrates HEM usefulness to predict real performance, but underscores the importance of quality input information to obtain these results. In this case there was uncertainty with model inputs, occupational patterns, weather data and HVAC system performance (i.e. heat-pumps and MVHR), and the high variability of these in real-world conditions.
We’re looking forward to more data from real-world trials of HEM, as it becomes established as the standard for domestic building modelling. This will continue to strengthen the robustness of the modelling methodology.
LETI response to Future Home Standard implementation of Home Energy Model: https://www.leti.uk/_files/ugd/252d09_aeb266fe44b542e2a00b9d47ad63b91c.pdf
HEM Inter-modal comparison: https://assets.publishing.service.gov.uk/media/6578b3da254aaa0010050bb3/hem-val-01-inter-model-comparison.pdf
HEM Camden Passivhaus study: https://assets.publishing.service.gov.uk/media/657859d90467eb000d55f5cd/hem-val-03-in-use-validation-camden-passivhaus.pdf
All HEM validation studies (including inter-model comparison, Camden Passivhaus, lab trials, and results from testing Building for 2050 and IEA projects): https://www.gov.uk/government/publications/home-energy-model-validation-documentation
✓ Desktop Access to HEM
✓ Documentation & Datasets
✓ Customer Support
✓ Try our full feature set
✓ Export data and reports
✓ Integrate with our APIs
