Sensor errors have an important impact on the operation, control, and detection of building energy systems. Correct and reliable sensors can effectively reduce the energy consumption of building energy systems. Virtual in-situ calibration (VIC) based on Bayesian inference and Markov Chain Monte Carlo method that no need to increase extra or install new sensors can effectively reduce systematic error and random error of the sensor and increase the reliability. However, due to the long calibration time, large amount of data, normalization constant and sensitivity coefficient required for the Whole calibration and Local calibration previously studied by the research group, all the sensors could not be fully calibrated or the calibration results were unstable. Based on the above problems, we proposed a new calibration method -- Component calibration, which divides the complex building energy system into several components to achieve accurate calibration without various coefficients, few data and calibration time. First, these three methods are applied to the actual primary air return reheating system, and the accuracy of Component calibration method is verified by comparing the results of the three calibration methods. Then, different results generated by different division of Local calibration domain are compared, which verifies from the side that Component calibration can perfectly calibrate systematic errors and random errors of all sensors.