A poorly calibrated model undermines confidence in the effectiveness of building energy simulation, impeding the widespread application of advanced energy conservation measures (ECMs). Striking a balance between information-gathering efforts and achieving sufficient model credibility is crucial but often obscured by ambiguities. To address this gap, we model and calibrate a test bed with different levels of information (LOI). Beginning with an initial model based on building geometry (LOI 1), we progressively introduce additional information, including nameplate information (LOI 2), envelope conductivity (LOI 3), zone infiltration rate (LOI 4), AHU fan power (LOI 5), and HVAC data (LOI 6). The models are evaluated for accuracy, consistency, and the robustness of their predictions. Our results indicate that adding more information for calibration leads to improved data fit. However, this improvement is not uniform across all observed outputs due to identifiability issues. Furthermore, for energy-saving analysis, adding more information can significantly affect the projected energy savings by up to two times. Nevertheless, for ECM ranking, models that did not meet ASHRAE 14 accuracy thresholds can yield correct retrofit decisions. These findings underscore equifinality in modeling complex building systems. Clearly, predictive accuracy is not synonymous with model credibility. Therefore, to balance efforts in information-gathering and model reliability, it is crucial to (1) determine the minimum level of information required for calibration compatible with its intended purpose and (2) calibrate models with information closely linked to all outputs of interest, particularly when simultaneous accuracy for multiple outputs is necessary.

The availability of the building's operation data and occupancy information has been crucial to support the evaluation of existing models and development of new data-driven approaches. This paper describes a comprehensive dataset consisting of indoor environmental conditions, Wi-Fi connected devices, energy consumption of end uses (i.e., HVAC, lighting, plug loads and fans), HVAC operations, and outdoor weather conditions collected through various heterogeneous sensors together with the ground truth occupant presence and count information for five rooms located in a university environment. The five rooms include two different-sized lecture rooms, an office space for administrative staff, an office space for researchers, and a library space accessible to all students. A total of 181 days of data was collected from all five rooms at a sampling resolution of 5 minutes. This dataset can be used for benchmarking and supporting data-driven approaches in the field of occupancy prediction and occupant behaviour modelling, building simulation and control, energy forecasting and various building analytics.

With the advance of the internet of things and building management system (BMS) in modern buildings, there is an opportunity of using the data to extend the use of building energy modeling (BEM) beyond the design phase. Potential applications include retrofit analysis, measurement and verification, and operations and controls. However, while BMS is collecting a vast amount of operation data, different suppliers and sensor installers typically apply their own customized or even random non-uniform rules to define the metadata, i.e., the point tags. This results in a need to interpret and manually map any BMS data before using it for energy analysis. The mapping process is labor-intensive, error-prone, and requires comprehensive prior knowledge. Additionally, BMS metadata typically has considerable variety and limited context information, limiting the applicability of existing interpreting methods. In this paper, we proposed a text mining framework to facilitate interpreting and mapping BMS points to EnergyPlus variables. The framework is based on unsupervised density-based clustering (DBSCAN) and a novel fuzzy string matching algorithm "X-gram". Therefore, it is generalizable among different buildings and naming conventions. We compare the proposed framework against commonly used baselines that include morphological analysis and widely used text mining techniques. Using two building cases from Singapore and two from the United States, we demonstrated that the framework outperformed baseline methods by 25.5%, with the measurement extraction F-measure of 87.2% and an average mapping accuracy of 91.4%.