Impact of measured data frequency on commercial building energy model calibration for retrofit analysis

Developing an accurate energy model remains challenging because of the numerous parameters that define a building's performance and the difficulty of the measuring them. Automated calibration using measured data can be used to develop an accurate energy model. This paper investigates the impact of the monitored data frequency (hourly vs. monthly) on the calibration results and retrofit analysis. A 11-storey government office building located in Ontario, Canada was selected as a case study to demonstrate the proposed methodology. Sensitivity analysis using a variance-based method was conducted to select the calibration parameters. The results of optimization calibration using two measured data frequencies demonstrated that monthly calibrations were unable to reflect actual operation conditions of a case-study building, thus indicating a necessity for hourly calibrations. Although the monthly calibrated model had the minimum average value of the CV(RMSE) of monthly energy consumption (7.4%), the CV(RMSE) of the hourly heating usage for that model was about 38.2%. Implementation of several energy saving measures on both calibrated models revealed that the resolution of measured data can significantly affect the estimated impact of energy saving measures.

Automated summary

Developing an accurate energy model remains challenging due to the complexity of building parameters and measurement difficulties. This study found that monthly calibration was unable to reflect actual building operation conditions, highlighting the necessity for hourly calibration. The resolution of measured data significantly affects the estimated impact of energy saving measures implemented on calibrated models.