Visit the Energy Systems Laboratory Homepage.
Comparison of Chiller Models for Use in Model-Based Fault Detection
Loading...
Date
2001
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Energy Systems Laboratory (http://esl.tamu.edu)
Texas A&M University (http://www.tamu.edu)
Texas A&M University (http://www.tamu.edu)
Abstract
Selecting the model is an important and essential step
in model based fault detection and diagnosis (FDD).
Factors that are considered in evaluating a model
include accuracy, training data requirements,
calibration effort, generality, and computational
requirements. The objective of this study was to
evaluate different modeling approaches for their
applicability to model based FDD of vapor
compression chillers.
Three different models were studied: the Gordon and
Ng Universal Chiller model (2nd generation) and a
modified version of the ASHRAE Primary Toolkit
model, which are both based on first principles, and
the DOE-2 chiller model, as implemented in
CoolToolsTM, which is empirical. The models were
compared in terms of their ability to reproduce the
observed performance of an older, centrifugal chiller
operating in a commercial office building and a
newer centrifugal chiller in a laboratory.
All three models displayed similar levels of accuracy.
Of the first principles models, the Gordon-Ng model
has the advantage of being linear in the parameters,
which allows more robust parameter estimation
methods to be used and facilitates estimation of the
uncertainty in the parameter values. The ASHRAE
Toolkit Model may have advantages when refrigerant
temperature measurements are also available. The
DOE-2 model can be expected to have advantages
when very limited data are available to calibrate the
model, as long as one of the previously identified
models in the CoolTools library matches the
performance of the chiller in question.