Applications of POD-based reduced order model to the rapid prediction of velocity and temperature in data centers

•Fast and simultaneous predictions of 3D velocity and temperature fields.•The average deviations between POD results and CFD data are within 0.5 °C.•Multi parameter POD-Galerkin projection and POD-insert method are both adopted.•Temperature distributions in rack and room scales of data center are analyzed. The predictions of fluid flow and temperature distributions in air-cooled data centers are crucial for improving the thermal management efficiency. In this work, the proper orthogonal decomposition (POD) method is adopted to rapidly and efficiently predict the three-dimensional (3D) fluid flow and temperature fields in data centers. The POD-Galerkin projection method is adopted to obtain POD coefficients of temperature while the POD-insert method is used to calculate POD coefficients of velocity. By comparing with CFD simulations, it shows that the POD method can well predict the fluid flow and temperature fields at the room scale, with an average error about 0.50℃. Besides, the computation time of POD-based reduced order model is approximately one thousandth of the CFD model.