Comparison of a whole-building and HVAC-system simulation model with measured data from a new office building

Abstract: Energy calculations are used for many purposes, for example during the design andconstruction phase to comply with Boverket’s building regulations, energy declarations or toachieve energy saving in a building. A main problem in the design process of a new building,is to accurately predict the energy performance. The energy calculations require data fromdifferent interacting components, which have been shown to be challenging to measure. Thisproblem arises in particularly for complex buildings, notably office buildings. Due to the lackof detailed energy models and analyses of the differences between the model and reality foroffice buildings in Sweden, this report aims to answer the question: What are the causes ofdeviations between the models and the measurements of energy usage and indoor airtemperature for the studied office building? The office building Sthlm New 04 was modeledas a whole-building within IDA ICE and the simulated data was compared with measureddata from 2021 that was provided by Skanska Fastigheter.During the design of the model, the simulation model was fed with external source files (withmeasured data) to make the model operate at the same time as the real system. Since therewas no logged occupancy data for this office building, it was assumed that occupancy wasdistributed equally in all office zones. The model has a detailed HVAC system to make themodel as accurate as possible.The result showed overall a good agreement with the measured data, especially for thesimulated district heating and cooling. However, a closer look revealed that there were somedifferences that the model did not account for. The building’s three Air-handling units alsoshowed a good agreement with an average of -0.3 °C, -0.2 °C and +0.2 °C. The resultshighlight several problems with modeling and measured data. Under the influence ofassumptions and hard to define input data, there may be misinterpretation by the modeler.Other reasons could be sensor errors or manufacturing errors by most of the errors could becorrected by a more detailed occupancy in the model. However, this study has shown thatthe detailed model can be used for a fair comparison between the simulated model and thereal building, although there may be some discrepancies. Undoubtedly, it is difficult to createa model that exactly matches the real building, but this model is a good representation thatcan be used for future research such as digital twin.