Virtual in-situ calibration method in building systems

• Calibration is mathematically defined using the framework of matrix and statistics.
• Calibration can be conducted automatically without moving sensors.
• Method is effective in reducing the systematic error due to changes in working conditions.
• Error of a supply air temperature sensor in RTUs is reduced from ± 19.2 °C to ± 0.7 °C.

Sensors play an important role in guiding building systems to achieve desired operation and efficiency. However, sensors are subject to continuous degradation and failures over time. Although a periodical calibration is needed, it is exceptionally difficult and/or impractical to many sensors with a conventional manual approach. Uncalibrated problematic sensors could significantly compromise the systems' performance and lead to unintended loss of energy efficiency in buildings. We propose a methodology, termed virtual in-situ calibration, to solve this critical issue. It is developed by mathematically extracting the characteristics of essential aspects involved in a calibration, including the environment assessment, benchmark establishment, and uncertainty quantification. A case study of a supply air temperature sensor in rooftop units illustrates the implementation process; the erratic uncertainty is reduced from ± 19.2 °C to ± 0.7 °C after the virtual in-situ calibration. The calibration method can be implemented online to significantly improve the reliability of sensor networks in buildings.