A Stochastic Model Predictive Control Strategy with Belief Reliability Constraints for Cloud Data Centers

Cloud data centers, as critical infrastructures, have control strategies that are crucial to ensure safe and reliable operation and efficient management. However, uncertainties that exist in actual operation can affect the robustness of control strategies. Most of the existing studies design control strategies based on a nominal model that ignores uncertainties, which is not robust enough in practical applications, i.e., it is easy to lead to performance fluctuations or even safety constraints violation. Although existing studies analyze the impact of uncertainty, they generally fail to incorporate active control uncertainty into the design, which makes it difficult to meet the robustness requirement especially when the uncertainty magnitude is large. To solve this problem, this paper introduces the belief reliability constraint to actively control the impact of uncertainty and proposes a stochastic model predictive control (SMPC) strategy for cloud data centers. The core lies in the establishment of a belief reliability model for cloud data centers, including the performance equation, margin equation, uncertainty quantification, and belief reliability measurement equation, and then the variables of performance, margin, and belief reliability are used as objective functions or constraints to construct the SMPC optimization model. The case study results show that the proposed method has advantages in terms of energy efficiency, robustness, and input fluctuation suppression compared to control strategies that do not actively control the impact of uncertainty. In addition, compared with no control strategy at all, the proposed method can reduce the electricity cost and carbon emission of cloud data centers by about 46%, which is both economically and environmentally friendly.