Hidden factors and handling strategy for accuracy of virtual in-situ sensor calibration in building energy systems: Sensitivity effect and reviving calibration

Abstract

Virtual in-situ calibration (VIC) can be conducted on a large scale, in-situ, to calibrate multiple working sensors in an operational building's energy system based on Bayesian inference. As well as random errors, the VIC can handle various systematic errors that are not covered by a conventional calibration, and it does not require removing working sensors or adding reference sensors as is done in a conventional calibration. For successful calibration under the various working conditions of a system, it is important to figure out hidden factors and their negative impacts on the accuracy of VIC. Through case studies for a LiBr-H2O refrigeration system, this study reveals two different sensitivity effects and how they affect the accuracy of VIC. Moreover, to handle the sensitivity issues, a new calibration strategy (named reviving calibration) is suggested and then evaluated in this work. This paper (1) shows the VIC problem formulation process, (2) explains how the two sensitivity effects influence the calibration accuracy, and (3) proves how and how much the suggested handling strategy solves the negative effects problem. The two case studies demonstrate the reviving calibration results in average 53% and 4% improvements, respectively, for temperature and mass flow rate sensors compared to the existing VIC method.