Virtual in-situ sensor calibration (VIC), using Bayesian MCMC with system models, can be applied to large-scale building sensor networks in order to calibrate multiple working sensors including virtual and physical sensors. Using the calibration strategies herein, VIC methods work to overcome various systematic errors (from simple to troublesome) that cannot otherwise be handled by a conventional calibration. Thus, it is important to understand and evaluate the VIC strategies in a whole system with various error conditions to ensure reliable performance of VIC in the operational stage. This study applied individual calibration strategies to a whole building system (a LiBr-H2O refrigeration system), and their effectiveness and limitations in solving the VIC problem were evaluated for the calibration of each type of working sensor under the various error conditions that cause negative effects. The combined strategies are also suggested to improve calibration accuracy by overcoming the individual limitations. This paper (1) explains how to formulate the individual and combined VIC strategies with Bayesian MCMC, (2) shows their capability to handle negative effects and limitations, and (3) demonstrates how combined strategies surmount the individual limitations for a high accuracy in various error conditions. (C) 2018 Elsevier B.V. All rights reserved.