Temperature field data are essential for intelligent environmental control systems. However, acquiring comprehensive temperature distributions in data centers with numerous sensors is not only costly but can also disrupt other equipment layouts. Thus, the capability to rapidly reconstruct temperature fields using a minimal number of sensors is crucial for smart adjustments in these systems. This paper introduces a genetic algorithm (GA) to the gappy proper orthogonal decomposition (POD) method, proposing an objective function to optimize the placement and quantity of sensors. This benchmark for optimization is applied to a case study on environmental control within a data center, where the optimal sensor layouts and the corresponding mean absolute errors were determined for configurations using 5 to 15 sensors, all maintained within 1 °C. Notably, the lowest reconstruction errors of 0.03 °C and 0.05 °C were achieved with a 12-sensor setup. Six uniform sensor layouts acted as control groups. Results demonstrated that error improvements in the optimized layouts were 98.25 % superior to those in uniform layouts. Moreover, transient CFD simulations were conducted to experimentally control the gappy POD algorithm. These simulations confirmed the capability of this approach to compute optimal air supply parameters and maintain temperature control within standard limits, resulting in a 24 % airflow savings compared to fixed ventilation modes under identical conditions. Consequently, this study effectively achieves rapid reconstruction and environmental control of temperature fields in data centers, offering important theoretical and practical insights for real-world engineering applications.
Read full abstract