Data Center Temperature Control using PI System and MATLAB

Several existing models are highly effective in predicting air temperature distributions of data centers. However, the complexity of these approaches requires significant computation time, making it difficult to control parameters in real time. Furthermore, lack of communication between inputs such as cooling infrastructure and outputs results in latency in receiving the relevant data and enacting changes. The goal of this case study is to demonstrate a simple heuristic model that enacts automatic controls to maintain a target server inlet air temperature by changing CRAC return air temperature set points. Data are collected and gathered with the PI System from OSIsoft, and are then imported into MATLAB, where regressions and conditionals are applied to generate a new optimal return air temperature set point. The new set point is written back into the PI system, which results in overwriting of the Building Management System via MODBUS communication. In addition, displays and notifications allow users for more effective monitoring of the conditions of the data center. The model developed from this case study represents an effective baseline for maintaining a target inlet air temperature for a server rack using real-time control techniques.