ISSN Online: 2377-424X
ISBN Print: 978-1-56700-474-8
ISBN Online: 978-1-56700-473-1
International Heat Transfer Conference 16
CONTROL-ORIENTED MODELLING AND EVALUATION FOR THE TEMPERATURE DISTRIBUTION IN DATA-CENTERS
Abstract
Energy efficiency optimization for data centers received wide attention in recent years. In order to optimize
the energy consumption of data center facilities via real-time control strategies, it is essential to establish
fast and accurate thermal predicting/evaluating models. Existing researches have proposed fast temperature
evaluation models (FTEMs) for data centers for steady-state flow pattern. The model parameters therein were
identified through CFD simulations. The main drawback of these models is that its accuracy will decrease
when the flow field is deviated from its designed state. The parameters corresponding to the new flow field
have to be re-identified through a set of CFD simulations. In practice, the flow rates of racks are usually being
tuned, with the goal of improving cooling efficiency and saving power. Hence, it is necessary to take different
flow patterns into consideration in the control-oriented data center thermal modeling. This paper proposed a
machine learning method to improve the FTEM. An artificial neural network (ANN) is constructed on top
of the FTEM. It learns the relationship between flow patterns and model parameters, and then it replaces the
time-consuming CFD-based parameter identifying process. Then, the temperature evaluation under different
flow patterns can be implemented by coordinating the ANN model and the FTEM. The accuracy of the ANN
based FTEM is validated by comparing with the pure CFD results. The proposed model can be used to design
real-time controllers for data centers with changing flow field.