Assessing uncertainty in co-heating tests: Calibrating a whole building steady state heat loss measurement method

Co-heating is a method of estimating the whole building heat loss coefficient (HLC) of a dwelling using constant internal temperatures and steady state analysis. Use of the co-heating method in the UK has provided significant evidence of a fabric performance gap and identified unexpected mechanisms for heat loss, such as the party wall bypass. However, to date there has been little assessment of the uncertainties associated with this method, leading to considerable debate and lack of understanding over its use. This research draws on the use of both simulated co-heating tests and case study field tests to understand uncertainty within the co-heating method. A broad range of uncertainties are assessed under three themes: weather driven, experimental and statistical uncertainties. For each source of uncertainty identified, the nature, direction and scale is considered. Interactions to key building characteristics are then explored, including the thermal mass, fabric insulation and airtightness of a test dwelling. In addition, approaches to both identifying and limiting these errors are discussed. In particular, the impact of the prevailing test weather conditions are shown to influence HLC estimates, particularly solar radiation. These include: the estimation of solar gains, the imperfect measurement of solar radiation, the influence of stored solar heating contributions and the influence of solar driven overheating restricting when reliable HLC estimates can be obtained. Furthermore, in non-airtight dwellings, the impact of wind is shown to increase variation in heat loss. Incomplete knowledge of secondary heat flows driven by the external environment lead to definitional uncertainty in HLC estimates and make comparisons to predicted or design HLCs more complex. Experimental uncertainties, from non-uniform internal temperatures, equipment measurement errors and uncoupled heat losses are also shown to potentially provide large systematic uncertainties if unchecked. Having established the presence and nature of these uncertainties the application of the co-heating method is reviewed. This includes suitable environmental testing conditions, the required duration for testing and the ability to perform comparisons to design and determine retrofit improvements. As such issues are a function of the building being tested and its characteristics, a number of archetype dwellings are used to show how the requirements and the general suitability of co-heating varies between dwellings. However, within a suitable external environment and avoiding experimental uncertainties, accurate HLC can be obtained with just 72 hours of monitoring. In addition, an approach to providing appropriate uncertainty estimates to a given co-heating test is developed and the interpretation of the measured HLC is shown to be when compared to both design predictions and when examining retrofit improvements. To summarise, theoretically, this research establishes the bounds of the co-heating method and demonstrates the effectiveness of co-heating tests in understanding building fabric heat loss. Methodologically, it establishes the role of simulation in the estimation of errors associated with measurement procedures and demonstrates the value of applying multi-method approaches to complex problems arising from the physical performance of buildings. Substantively, this research highlights the need for researchers working in the field to be mindful of the uncertainty in co-heating tests and understand limits of the measurement and its interpretation.