Virtual sensor-assisted in situ sensor calibration in operational HVAC systems

A sensing environment is important for the reliable control and performance of operational heating, ventilation, and air-conditioning systems. Virtual in situ sensor calibration (VIC) has been developed to estimate various sensor errors, especially systematic errors caused by working environments, to correct erroneous measurements in the field. However, it is difficult to apply VIC to operational systems, because key sensors for the VIC formulation are not usually installed in real systems. This makes VIC impossible or the problem difficult to determine accurately. To improve its applicability and performance by handling this practical challenge, an advanced in situ sensor calibration method that uses virtual sensors for replacement and backup purposes is proposed. The virtual sensors for replacement make it possible to construct the calibration formulation, even with a limited number of physical sensors. The backup virtual sensors can reinforce the formulation mathematically, thus improving calibration accuracy. In a case study for a typical air-handling unit, the suggested virtual sensor-assisted VIC was constructed based on the model-driven virtual sensors, and it showed excellent calibration accuracies for multiple simultaneous sensor errors by adding data-driven virtual sensors. In addition, with the various virtual sensors developed, the use of multiple operational datasets was effective in calibrating simultaneous errors. The calibration error decreased from 72.7% to 5.2% using the virtual sensors.