6. Single versus multi zone models: simulation study
6.1.3 Parameters and hypotheses
The overall hypothesis is that shifting from a single-zone model with fixed equivalent set-point temperature (as considered in Belgium) to a single-zone
model with correction factors for intermittent and spatially reduced heating (as considered in the Netherlands and Germany) and further to a multi-zone model can result in a significant reduction of the discrepancies between real and theoretical values. The single and multi-zone approaches presented in Chapter 5 are different with regard to which aspect they can or cannot take into account.
Some are considered to explain part of the prediction error described in literature and in the previous chapters, being characterised by an average overestimation of the energy use that increases at low performance levels (see previous section) and also by a large spread in prediction errors at all performance levels when comparing values at the level of the individual house and household.
Chapter 3 puts forward the hypothesis that part of the large overestimation of the energy use in old houses results from the fact that, by considering a fixed equivalent set-point temperature of 18°C, the official calculation method that was used overestimates, on average and at building level the average indoor (5.3.2) should result in more accurate calculated indoor temperatures and space heating demands. However, this will probably not explain the full extent of the prediction gap. Firstly, Chapter 5 (5.2.2) showed that the equivalent set-point temperature resulting from the German and Dutch approach do not always result in values below 18°C. Secondly, large overestimations of the real energy use by regulatory energy calculation models were found not only in Belgium, but also in countries where the single-zone calculation methods include these types of correction factors: Germany, the Netherlands, the UK [29,32,97].
Therefore, a second hypothesis from Chapter 3 is analysed: considering the real window opening profiles should result in lower calculated air flow rates than considered in the standard EPB-model and therefore in lower heat losses, a lower theoretical energy demand and a smaller prediction gap (3.4.1).
Considering the heating profiles and window opening profiles together leads to the third hypothesis from Chapter 3. The windows that were opened by the inhabitants were mainly those of the unheated bedrooms and of the bathrooms, but not those of the heated living rooms. Therefore, single-zone modelling approaches, which do not consider this behavioural relation between ventilating and heating at room level, will overestimate the energy demands even when realistic building average temperatures and total air flows are considered (3.4.1).
Taking this zonal differentiation into account in the multi-zone model should thus result in smaller overestimations.
Two additional differences between rooms are not taken into account in single-zone models with correction formulas: the different internal heat gains and the different performance levels of the external envelope of each room. In a similar way as for the ventilation heat losses, all internal heat gains of a house are summed up in single-zone models, without considering if those heat gains
originated in the directly or in the indirectly heated areas. However, the most heated area is the living area, where cooking takes place and where people are active, using electrical appliances (e.g. televisions) and, especially during winter when nights are long, requiring active lighting. Taking into consideration this zonal match between heat gains and thermal comfort requirements in a multi-zone model may remove one more part of the prediction gap. Similarly, while corrected for taking into account the fact that not all rooms are heated, the single-zone approaches from DIN 18599 and NEN 7120 still simplify the thermal properties of the envelope into one average heat loss coefficient and the spatial differentiation into one single equivalent set-point temperature. This simplification can be compared to the assumption of a homogeneous occurrence of heat losses over the total heat loss area. However, walls, windows, roofs, floors etc. have different thermal properties and are not distributed uniformly over the whole building envelope. Loga et al. [175] reported that making different modelling assumptions on which zones of their model were heated and which were not, influenced the correlation between the unheated area fraction and the correction factor for spatially reduced heating (see 5.2.2). Using a multi-zone model of the considered house instead of a single-multi-zone model with correction factor based on other housing typologies and zonal heating patterns should thus reduce the errors on the individual case-level. Moreover, it could also reduce the average overestimation of the energy use in not insulated houses because of the common lay-out of houses having their heated living area on the ground floor and their unheated night zone on the first floor. In a house where walls, roofs and floors are not insulated, the uninsulated floor will have the lowest equivalent thermal transmittance because of the thermal resistance and buffering capacity offered by the ground.
Another zonal differentiation that is not considered in the regulatory performance assessment models does not regard the different user profiles in adjacent rooms of the house, but the different user profiles in adjacent houses. In a simulation based study on apartment blocks by Nielsen And Rose [183], lowering the temperature of an apartment by 2°C was found to increase the energy demand of the adjacent apartments by 10 to 20%. Apartments are more compact than single family houses and therefore the relative effect of heat losses through party walls on their energy demand is supposedly higher, but a statistical association found in Chapter 3 gives sufficient ground for a small additional analysis using the multi-zone model. Chapter 3 showed that part of the variation in real energy use found for the (inhabited) houses of the old neighbourhood (‘cs1’) was associated with the fact some of the adjacent houses were not inhabited, thus causing heat losses through the party walls while the assessment method does not consider heat losses through those party walls. The large variations in heating profiles discussed in the first chapters makes it likely that not considering any heat transfer through party walls could also cause significant prediction errors if the adjacent houses are inhabited even though not necessarily biased errors leading to systematic under or overestimations.
One last parameter that is analysed is how the inter-zonal air flows through doors affect the simulation results. Loga et al. [175] referred to the errors that could be caused by uncertainties regarding the inter-zonal heat transfer because of the
uncertainty regarding the opening and closing of doors by the inhabitants. ISO 13790 [20] also mentions this additional complexity of considering inter-zonal air flows when shifting from single-zone to multi-zone models. Furthermore, Chapter 3 found that some living rooms were heated also at night or at high temperature and put forward the hypothesis made also by Weihl and Gladhart [142] that these types of behaviour are found to be used in uninsulated houses with local heating systems as a necessary solution for reaching bearable temperatures in the coldest spots. This puts the use of doors in the list of options a user has for conditioning different zones of the house. A sensitivity analysis is therefore included to see to what extent assumptions on the door opening profiles influence the calculated indoor temperatures, the theoretical energy use and the associated temperature take-back.