Virtual sensors use data from low-cost measurements and calibrated models to provide outputs that would either be too expensive or impossible to measure directly. Virtual sensor technology has the potential to enable cost-effective implementation of advanced monitoring, diagnostic, and/or control features for buildings. While it is commonly known that the reliability of virtual sensors depends on the amount and conditions of calibration data, no methods have been presented that quantify the effect of the conditions of calibration data on virtual sensor output uncertainty. In this paper, a general method is presented for estimating the virtual sensor output uncertainty in terms of the uncertainty, conditions and amount of calibration data. The method is demonstrated with a power consumption virtual sensor for packaged air conditioning systems.
- Virtual sensor; Uncertainty; Leverage; Regression diagnostics; Packaged air conditioners
Available at: http://works.bepress.com/howard_cheung/18/