Rapid Temperature Predictions in Data Centers using Multi-Parameter Proper Orthogonal Decomposition

Abstract

A proper orthogonal decomposition (POD)-based multi-parameter, reduced-order modeling framework that rapidly predicts air temperatures in an air-cooled data center is developed. The modeling parameters are heat load and time. The framework is applied on initial temperature snapshots acquired near a server simulator rack by measurements at discrete time instants and selected rack heat loads. To estimate the accuracy of the modeling framework, the predicted temperature data are compared with corresponding experimental observations. The proposed algorithm is demonstrated to be effective and efficient for full-factorial parametric temperature characterization.