Numerical and compact models to predict the transient behavior of cross-flow heat exchangers in data center applications

Abstract

Hybrid air/liquid cooling systems used in data centers enable localized, on-demand cooling, or “smart cooling” using various approaches such as rear door heat exchangers, overhead cooling systems and in row cooling systems. These systems offer the potential to achieve higher energy efficiency by providing local cooling only when it is needed, thereby reducing the overprovisioning that is endemic to traditional systems. At the heart of all hybrid cooling systems is an air to liquid cross flow heat exchanger which regulates the amount of cooling that the system provides by modulating the liquid or air flows or temperatures. Understanding the transient response of the heat exchanger is crucial for the precise control of the system. The aim of this work is the development of a rear door heat exchanger compact model using Artificial Neural Networks (ANN). The transient behavior of the heat exchanger is studied using a Finite Difference (FD) model. Different temperatures perturbations are introduced in the heat exchanger model to study its transient response. The finite different results are then used to train an ANN compact model. Both models are compared in terms of accuracy and computational resources.