Assessment of Building Energy Performance through Grey Box Modelling

Being able to confirm that the thermal performance of a building envelope is achievable and achieved, is a challenge faced by the construction industry. In that context, UCLouvain is developing methods, in collaboration with Jacques Delens s.a., to control the thermal quality of the building envelope in the construction phase, through a 9 days monitoring period without occupancy nor ventilation. The tests are performed in apartment units. The building envelope thermal performance is characterized by its Heat Transfer Coefficient, HTC, expressed in W/K. A difficulty is related to the presence of common walls and floors shared by adjacent units as it is not possible to gain access to all of the surrounding apartments or spaces when undertaking the tests. The difficulty is overcome by the placement of internal insulation on party walls, focusing on the determination of the overall heat loss coefficient of the external facades. Such experiments were performed in summer 2017 on two pair of south oriented identical units. Two types of boundary conditions were tested: the shared walls of two identical units were covered with insulation panels, while the shared walls of another pair of units were left uncovered. Two types of solicitations were tested as well: within each pair of identical units, one unit was left in free floating mode while the other was submitted to co-heating. A two-step methodology is then adopted for the data analysis. A reference HTC, not affected by solar heat gains uncertainties, is obtained by comparing the results of co-heating and floating temperature mode on identical units. The analysis of the data collected on each separate unit is then performed in order to provide values of the HTC that are expected to confirm the reference HTC. The analysis of the heat balance of each unit separately requires a reliable estimation of the solar heat gains, including the reflections of the surrounding masks and the shading due to the screens surrounding the windows. While a stationary analysis using a simple linear regression can be considered in some cases, the dynamic effect due to the variation of the internal temperature set points can be observed in most of them. A dynamic analysis is undertaken using a grey-box modelling approach with a forward selection procedure. The Root Mean Square of the Error between the measured and the calculated indoor temperature profiles is used as an indicator of the quality of the model. The probable uncertainty of the HTC is calculated on the basis of the chi-square merit function. For the tested units that are covered by inside insulation panels, the reference HTC obtained by dynamic analysis is almost equal to the predicted HTC (7.4 W/K). The confirmed mean HTC values obtained by dynamic approach show a maximum gap of -1 W/K. This gap increases up to 3 W/K when the analysis is performed trough a stationary approach. When the tested units are not covered by inside insulation panels, the mean HTC values obtained by either stationary or dynamic approaches confirm the predicted HTC (6.6 W/K), provided the test is performed in floating temperature mode. The HTC values obtained by a co-heating test at 29°C show much larger discrepancies ranging from 4 to 4.4 W/K depending on the analysis method. Those discrepancies are due to the presence of internal thermal bridges whose effect is magnified by the co-heating experiment. The probable uncertainty associated to the HTC is much higher when the tested units are not covered by inside insulation panels, whatever the testing method (free floating temperature or co-heating) or the analysis method (stationary or dynamic). The relevancy of the placement of inside insulation is assessed considering the AIE Annex 71 matrix: the cost of such a placement can be assessed in terms of labor cost, as well as its impact in terms of accuracy of the results