Cooling Capacity Prediction for Safety-Guaranteed Optimization in IoT-Enabled Data Center

The increasing demand for computation and storage has led to a significant rise in energy consumption and carbon emissions in data centers. Optimizing cooling systems to reduce Power Usage Effectiveness (PUE) has become a key concern due to its significant impact on the energy consumption of data centers. Thanks to the development of Internet of Things (IoT) techniques, previous studies have employed machine learning methods, which are representative data-driven techniques that are superior to traditional methods, to predict the energy consumption of the data center for assisting optimization. However, these methods have not addressed the safety issues that may arise during the optimization process. In this paper, we investigate the problem of predicting cooling capacity to ensure safety-guaranteed PUE optimization in IoT-enabled data center by leveraging machine learning. Specifically, we develop a safety-guaranteed PUE optimization framework for IoT-enabled data center, and utilize a neural network to predict the cooling capacity, ensuring safety during the optimization process. To further reduce the complexity of prediction models and improve their performance, we propose a general supervised autoencoder for dimension reduction and feature extraction. Extensive experiments demonstrate the effectiveness and superiority of the proposed neural network and supervised autoencoder over other machine learning methods.