Real-time Energy Management of Large-scale Data Centers: A Model Predictive Control Approach

With the development of cloud computing and 5G communication, the power load of data centers has increased rapidly. While bringing a huge burden on the environment, high electricity bills also put heavy economic pressure on cloud service providers. In order to improve the economic efficiency of enterprises, data centers need energy management. This paper proposes a real-time energy management method based on Model Predictive Control (MPC) for large-scale data centers powered by renewable energy. In this work, the energy consumption model integrates renewable energy, dynamic electricity prices, battery, TES tank, delayed execution of batch jobs, etc. In addition, Monte Carlo simulation is used to deal with the uncertainty of renewable energy output, workload and electricity prices. The proposed MPC algorithm can use predictive data and forecast error distribution to perform real-time energy management on data centers. Finally, case study demonstrates the effectiveness of the proposed MPC algorithm in real-time energy management of data centers.