Research On Data Center Energy Management Optimization Based On Data-Power Collaboration

In recent years,with the booming development of cloud computing and the arrival of the digital economy era,data centers have been developing rapidly in the global scope.With the increasing number and scale of data centers,the problems of huge energy consumption and high operating cost are prominent,which sets off the research upsurge of data center energy management.Different from traditional power users,the time-space transfer capability of data center workloads not only has significant advantages in improving energy utilization efficiency,but also has absolute advantages in responding to power system load regulation needs.Furthermore,in the electricity market environment,the spatial difference and temporal fluctuation of electricity prices also bring huge optimization space for data center electricity costs.Therefore,based on the temporal and spatial transfer characteristics of data center workloads,and combined with the temporal and spatial differences of electricity prices,this paper conducts research on the optimization of data center energy management based on data-power collaboration.Firstly,the basic factors of data center energy management are analyzed and studied.Through the analysis of the data center energy consumption,workload characteristics,power supply situation and the operation mechanism of the balance market,this paper discusses the time and space adjustment ability of the data center power load,the coordinated dispatching ability of the workload and power supply under multiple power sources,and the impact of the participation in the balancing market on data center energy management,and then explore the breakthrough point of the research.Secondly,research on the optimization of node-level data center energy management is carried out,and the main principles of node-level data center energy management with multiple power sources are proposed.According to these principles,a single data center energy management model based on uninterruptible power supply(UPS)nodes was established.Considering the UPS performance inside the data center and the electricity price levels in different periods,the workload was optimized in the time dimension and the UPS node dimension,and the appropriate power supply mode was selected,to achieve the goal of reducing energy consumption and electricity costs.The optimal scheduling decision is obtained by solving the mixed integer linear programming problem,and a comparative example is set to verify the effectiveness of the model.Finally,from a single data center to a data center cluster,the energy management of a data center cluster distributed across domains is studied.Under the condition that the data center cluster participates in the cross-domain power balancing market,based on the prediction of the balance market regulating direction and regulating price,the time and space transfer characteristics of the data center workload and the complementarity of the balance resource demand of different regional power markets are combined,and the data center cluster energy management model driven by the balancing market is established.Considering the uncertainty and prediction error in the optimization process,a two-layer stochastic receding horizon optimization algorithm is proposed to solve the model.The actual data is used to analyze the calculation example,which proves the ability of the data center to participate in the balance market,and obtains the optimal scheduling decision for the overall benefit of the cluster.Through the comparative analysis of various scenarios,it is proved that the data center cluster energy management driven by the balanced market can realize the coordinated optimization of data and power,which can bring additional benefits to the data center and reduce electricity expenses while promoting the balance of the power system.