A method to calculate uncertainty of empirical compressor maps with the consideration of extrapolation effect and choice of training data

Although compressor maps are powerful tools to estimate compressor power consumption quickly, their ability to extrapolate outside their training data range is always questioned. In order to quantify the effect of extrapolation, a method to calculate the uncertainty of compressor map outputs is developed in the current article. The method considers four major components of uncertainties due to various sources such as measurement uncertainties of training data and equation of state. The change of the predicted map uncertainty with the degree of extrapolation and the map accuracy are shown for eight different compressor maps. The results indicate that the uncertainty from model random error increases significantly as the maps extrapolate though extrapolation does not necessarily imply inaccurate compressor map outputs. The results also show that the uncertainty due to training data is the most significant component of uncertainties when the maps are not extrapolated.