Comparative validation and empirical calibration of a whole building simulation software – focus on modeling constraints

Dynamic simulation software is a useful tool for predicting various characteristics of building operation, such as energy consumption for heating and cooling, internal temperatures, and solar and heat transfer. However, a key issue is the reliability and validity of simulation results. Building modeling should be preceded by validation and calibration procedures that help adjust model parameters to ensure simulation results are as close to reality as possible. This article presents one validation and one calibration example of a commercially available simulation software, BSim, based on the control volume method. These cases use different reference objects and weather datasets, resulting in various modeling challenges and different sources of uncertainty and error. The first validation case relied on a relatively simple configuration specified in a European standard. In this case, the primary issues were related to minor gaps in the test specifications and the provision of key solar radiation climate data in a format that could not be processed by the software. Then, calibration was performed based on measurements in a free-running climatic chamber at the Rzeszów University of Technology, Poland. Here, the problems were attributed to the lack of information about the properties of building materials and the insufficient scope of measurements of the components of solar radiation. Sensitivity analysis showed that the model was most affected by changes in the thermal capacity of the outer shell, resulting in a nearly linear increase in all error indices (MBE, MAE, RMSE, and CvRMSE), indicating deterioration in model accuracy. In contrast, changes in thermal resistance primarily influenced the systematic error (MBE), while having only a minor impact on MAE, RMSE, and CvRMSE. Results also implied that moderate time-step refinement is sufficient to achieve reliable simulation accuracy while reducing computational cost.