Calibrating an archetype model for predicting space heating energy consumption in older residential buildings: a case study of a building in Oslo and Trondheim

Abstract

This thesis is part of the COFACTOR project led by the SINTEF Community. The project aims to identify when and why peak loads occur through the collection of energy measurements of buildings. The thesis is related to a specific work package within the project with a narrow scope, where the focus is how we can use building simulations as a tool to estimate the hourly power needs of buildings, as well as peak power consumption. The thesis aims to develop an archetype model for energy simulations by calibrating an existing model to fit measured data from case building in Oslo.

Building energy simulations are essential in Norway's construction sector, and archetype models provide insights into typical energy usage. Archetype models in building energy simulations are derived from building typologies, enabling estimates of energy use while reducing computational cost. The TABULA project created a building archetype through building typologies.

The thesis focuses on older residential apartment blocks with district heating, as they represent the largest share of measurements. A representative building from two different housing cooperatives was chosen. The archetype model is calibrated using parametric runs in IDA ICE. SINTEF collected energy measurements from buildings in Norway and meteorological data from in close vicinity of the buildings.

Calibration of archetype models towards actual measured data is necessary for accuracy, and the ASHRAE Guideline 14, IPMVP, and FEMP guidelines for validation as calibration studies predominantly utilize CVRMSE and NMBE to assess model calibration.

In conclusion, the thesis finds that the baseline model generally over-predicts energy consumption for space heating compared to measured data. Reducing the power capacity of the radiators has a significant impact on the model's prediction of energy consumption for space heating. Calibration of archetype AB\_02 for measured data for space heating from Building\_6488 shows that it is possible to make a good model with IDA ICE as a calibration tool. The calibration iteration with the best fit, Run10423, had an hourly CVRMSE, NMBE and R$^2$ of 16.62\%, -0.18\% and 0.88, respectively. GOF is 11.75, implicating a good fit for the model. Evaluating the model by testing an out-of-pool sample of a representative building from a different housing cooperative gave an hourly CVRMSE, NMBE, and R$^2$ of 29.59\%, -7.65\%, and 0.55, respectively. Compared to the uncalibrated model this is an improvement in CVRMSE of 22.82\%. For NMBE, the calibrated model is worse than the uncalibrated model by 5.86\%. The calibrated model generally underpredicts the energy consumption, as can be seen by NMBE of -7.45\%. This indicates that throttling the design power in the radiators and lowering the setpoint is overly calibrated to fit the energy consumption of Building\_6488.