4.2 Work Package 2: Evaluating Modelling Uncertainty in the Simulation of ICF in Whole
4.2.3 Modelling Uncertainties in ICF Simulation
The analysis started by looking at the annual and hourly simulation results provided by the two BPS tools when the user relies on the default settings. The monthly breakdown of annual and peak heating and cooling results showed a high discrepancy between the two BPS tools, particularly in the simulation of annual heating demand for the ICF building (up to NRMSE = 26.05%) (Step 0 of Fig.4.7). Among the three construction methods, the ICF and HTM cases showed the largest discrepancies, indicating that the amount of thermal mass in the fabric affected significantly disparity of results. The LTM building showed better consistency in comparison to the other two construction types in both annual and peak heating and cooling demand.
The monthly simulation results provided by the two BPS tools for the heating and cooling demand (Fig.4.7 and Fig.4.8, respectively) showed the largest discrepancies over the winter months, when the solar angle is small, for all three construction methods. In contrast, a relatively good agreement was achieved during summer. This suggested that further investigation was required to address the differences in the way the two BPS tools simulate solar gains.
Figure 4.7 “Equivalencing” the models. Monthly breakdown of annual heating energy predictions provided by tool E and tool I for all three constructions: (a) ICF, (b) low thermal mass (LTM) and (c) high
Figure 4.8 “Equivalencing” the models. Monthly breakdown of annual cooling energy predictions provided by tool E and tool I for all three constructions: (a) ICF, (b) low thermal mass (LTM) and (c) high
Furthermore, it was attempted to get a better understanding on the impact of default settings and solution algorithms on the dynamic performance of ICF and thermal mass. For this purpose, hourly simulation results were analysed for the internal, intra-fabric and external wall surface temperatures and for the heating and cooling demand for three consecutive days in the winter and summer periods. Differences in the hourly predictions of cooling demand were negligible (Fig.4.9), whereas the hourly results for heating demand showed that the largest disparity was again observed in the ICF simulation (Fig.4.10).
Figure 4.9 Hourly breakdown of cooling demand. Simulation predictions provided by tool E and I for three consecutive days in the cooling season (26–28 July) for all three constructions: (a) ICF, (b) low thermal mass (LTM) and (c) high thermal mass (HTM), when the user relies on the tools' default
settings
Figure 4.10 Hourly breakdown of heating demand. Simulation predictions provided by tool E and I for three consecutive days in the heating season (03–05 January) for all three constructions: (a) ICF, (b) low
thermal mass (LTM) and (c) high thermal mass (HTM), when the user relies on the tools' default settings.
Simulation predictions from both tools for the hourly wall surface temperatures showed a relatively good agreement (in terms of relative differences) for all three constructions (with the exception of outside surface temperatures). However, the absolute divergence indicated instances of maximum difference as high as 5oC (i.e. internal surface temperature of ICF building - Fig.4.11). This highlighted that the selection of BPS tools could significantly affect the outcome of thermal comfort assessments and could result in different conclusions being drawn about the thermal performance of the building.
Figure 4.12 Hourly breakdown of the inside surface, intra-fabric and outside surface temperature of the east wall. Simulation predictions provided by tool E and I for three consecutive days in the cooling season (26–28 July) for all three constructions: (a) ICF, (b) low thermal mass (LTM) and (c) high
thermal mass (HTM), when the user relies on the tools' default settings.
In a process of making the models equivalent for comparison, identical algorithms and input values were specified in both BPS tools. The impact of each algorithm that was investigated as part of the “equivalencing” process is analysed in detail in Section 3.2.1 of Appendix B. Moreover, Figs. 4.7 and Fig.4.8 show the results of the two tools for annual heating and cooling demand for each step of the process. The general observation was that the two most influential parameters leading to discrepancies in results were:
• the distribution of direct solar radiation
• the specification of surface convection coefficients.
Following the model “equivalencing” process, the annual simulation predictions provided by the two BPS tools were much more consistent for all three construction methods, with the exception of annual cooling demand for the ICF building (as illustrated in the black bars in Fig.4.13 below). However, the divergence in the prediction of annual cooling demand of ICF increased after the “equivalencing” process. This showed that there is a level of modelling uncertainty allied to ICF simulation that requires further investigation through measurements and empirical validation. The hourly simulation results provided by the two tools for the “equivalenced” models also showed some negligible inconsistencies in terms of both absolute and relative differences, as presented in Section 3.2.3 of Appendix B.
Figure 4.13 Absolute difference and NRMSE between the simulation predictions provided by tools E and I for the three construction methods ICF, low thermal mass (LTM) and high thermal mass(HTM), when
During the “equivalencing” process, several observations were made on how the different modelling methods employed by the tools affected the results’ discrepancy even when the input values were the same (in this case the climate data). As a result, two modelling factors were analysed:
1. The solar timing (used in the calculation of the solar data).
2. The impact of variations in wind speed (for the calculation of the external surface convection coefficients).
This analysis is presented in Section 3.3 of Appendix B. The general conclusion was that the variation observed in the simulation predictions was higher for heating demand and increased according to the level of thermal mass in the fabric. Consequently, the most profound inconsistencies were observed once again in the simulation of the ICF and HTM buildings.