Safety Ensured Machine Learning Framework for Energy System Operation in Data Centers

With the increasing demand for web-based services and the growing amount of data generated and stored, data centers are becoming more vital for organizations of all sizes. Unfortunately, climate and abrupt weather changes can compromise data center conditions and cause costly and inconvenient performance issues. Therefore, it is imperative to maintain appropriate environmental conditions within data centers to ensure the equipment’s safety and efficiency, as well as the comfort and safety of the workers. Monitoring and managing the temperature and relative humidity within a data center can minimize the risk of equipment failure and employee discomfort, ensuring the reliability and productivity of the facility. This paper examines machine-learning (ML) models’ applicability for handling near-real-time emergency airflow control in data centers during extreme weather events. ML models are built in Python and evaluated using mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination $R^{2}$ metrics. In addition, energy plus simulates the environment of a data center. Finally, the data center cooler settings, IT equipment load level, and outdoor air conditions are used to manage the facility’s requirements. The seventh-order polynomial model had the highest $R^{2}$ score and the lowest errors during training and testing.