OR-20-C071 -- Method for Tuning a Chiller Power Estimation Model with a Very Limited Part Load Performance Data Set PDF

This paper is focused on testing a recently developed chiller power estimation method, which is based on understanding the processes nature and building each term in the equations system with intention to reflect some distinct phenomena and to identify the meaning of its value. The method is tested when only very limited part load performance data is available for determining the equations’ coefficients. Extrapolation of a five-points data set, calculated with the new developed method, is successfully verified for a centrifugal one-compressor one-stage variable speed chiller. High accuracy is achieved in the extrapolated data. Successful extrapolation of such a reduced data set confirms the accuracy of the developed method is not accidental, because the model has a theoretical basis and each term as well as all coefficients reflect some physical phenomena and the terms behavior are clearly understandable and predictable.

To better recognize the developed method benefits and advantages, it’s compared with another method which has been widely used for many years, and which is based on an empirical bi-polynomial equation with six to ten terms. For comparison, the method is tested using the same five part-load data points of the same chiller, plus one more required to calculate the six coefficients of the equation.

It is found the new method is accurate for interpolation and extrapolation, while the old method can be used for interpolation, but it is not suitable for extrapolation, as the errors of estimated power for points which are not located close to the base six-points curve are unacceptable.

The new method has obvious advantages as it requires a smaller data set for determining its coefficient and can be used for accurate data extrapolation. Therefore, it is more effective and versatile than the old commonly used chiller power estimation technique.