Case Study

Technology Validated in Relevant Environment

This document describes a series of ‘real-world’ tests carried out on a high spec, new-build dwelling. The overall results are summarized and discussed below.
Case Study Timeline

1. Case Study Objective

Validate our (patent applied for) method in a real building, rather than an idealised insulated box.

Expected difficulties during transition to real building include:

  • Variable outdoor temperatures
  • Wind and rain
  • Effect of floor/ground modelling, as not a straightforward thermal resistance to the outside.

2. Validation means:

  • Same building -> same result
  • Change something in the building -> result changes
  • Ideally, home passes as it was specified to do so, and is believed to be ‘to specification’

3. Key questions to answer

  • Are the measurements repeatable?

Yes, although it took a few shots to establish the procedure we got consistent results between tests, estimating heat loss to the outside within 10% (test 3) to 13% (test 5).

  • Can we show that a building meets specification with a simple screening test?

Yes, the new building passes as expected.

  • Do measured and predicted values of heat loss match?

Yes, we have hand calculated for the test building and found an agreement within 3%.

  • Have we shown that removing insulation from a building makes a difference to heat loss?

Yes, removing 6m2 of loft insulation increased measured heat loss by around 14W/K (expected value 12-15W/K)

  • Did the measurement setup work well?

Data format is robust to sensor dropouts, but Z-wave sensor communication did not prove wholly reliable.

4. Test set-up

5. Example of Equipment Conditions , Timings & Locations

6. Findings

  • Reproducibility: The methodology is reproducible – similar test parameters lead to similar results despite different external conditions.
  • Building Analysis: The method produces results which were consistent with the building specification. Removing insulation led to consistent change in measured insulation.
  • Wind Sensitivity: We are currently unable to estimate impact of wind on the consistency and reliability of the results.
  • Limit Pre-Heating: Limited pre-heating improves the results, however when carried to excess (Test 2) it caused unreliable results. For valid results, allow the effects of pre-heating to dissipate before the heating & cooling phases of the experiment.

7. Reproducibility


  • Comparing the results from test 1 and test 2a (using the same duration of heating) leads to very similar results
  • Test 2b, 3 and 5 all produce consistent estimates of the buildings specific heat capacity
  • Conditions for tests 2b, 3 and 5 were different, so differing K-values are expected.

8. Comparison with building specifications

– Building analysis: The method produced results which were consistent with the building specification. Evidence:
Test 3 estimated K-values close to the building specification – error bounds contained the building specification, this implies the building may have slightly worse insulation than specification, but nor enough for the method to be conclusive.
– Estimated heat capacity in tests 2b, 3 & 5 are very consistent with building specification, suggesting we have obtained valid results using 5 hours of heating.
– Test 2b & 5 estimated K-values slightly worse than the building specification – 2b had higher wind conditions & excessive pre-heating and test 5 had insulation deliberately removed.

9. Additional Learnings

– Narrower Bounds: The method returns lower and upper bounds for the K-value. These were based on some initial assumptions about the temperature sensor biases and the quality of air mixing achieved. Narrowing these bounds increase the usefulness of the test.
– Test 5 has produced some very good data allowing both of these to be re-assessed. For these tests tighter bounds can be generated and are shown in red in the results table below.
For Test 5, where we removed 6m squared of loft insulation, these new bounds change the conclusion; it is now that the building is not performing to specification.