A study on the sensor calibration method using data-driven prediction in VAV terminal unit

In this paper, sensor calibration method using data-driven prediction models are developed to eliminate sensor error in the variable air volume (VAV) system. Indoor sensible loads and indoor carbon dioxide (CO2) concentrations, which are the main factors necessary for sensor operation in a VAV system, were predicted using a data-driven model (artificial neural network). Using this prediction model, we developed a method to calibrate the sensor error by using it to derive a system model of the sensor that we want to calibrate. As a result of the performance evaluation of the indoor sensible load prediction model, MBE was -1.8% and Cv(RMSE) was 3.4%. The performance evaluation of CO2 prediction models showed that MBE was -3.2% and Cv(RMSE) was 5.4%. In verification of sensor calibration method, it was confirmed that the error occurring in the sensor can be corrected through the application and verification of the method and procedure for calibrating the VAV terminal unit sensor using the prediction model. In addition, it was confirmed that the error can be corrected both in the case of a single error (CASE1 ti CASE3) in which an error occurs in one sensor and in the case of two or more multiple errors (CASE4 ti CASE7). In addition, the calibration of the sensor data was able to solve practical difficulties such as sensor replacement. (C) 2021 Elsevier B.V. All rights reserved.

Automated summary

A sensor calibration method using data-driven prediction models was developed to eliminate sensor errors in VAV systems. Performance evaluations showed the method effectively corrected errors and could handle both single and multiple errors. The calibration of sensor data also addressed practical challenges such as sensor replacement.