Research On Demand Response Strategy Considering The Adjustment Potential Of Data Center Load

The construction of novel power system has higher requirements for the capacity of power supply-demand balance and control means of the power grid.Establishing an adjustable load resource pool,including data centers,is the major trend of future development.However,the adjustment potential of the data center load has not yet been accurately quantified,as extensive internal and external factors affect the reduction,spatial transfer,and temporal transfer capabilities of loads,which makes the data center cannot effectively participate in demand response programs.Ignoring the adjustment potential of the data center load and its’influencing factors could cause a situation where the load optimization plan made by the data center would deviate from the actual operating conditions.This situation will decrease the expected revenue of the data center participating in demand response and even affect the main business of the data center,which can cause unnecessary losses and extra costs.Therefore,this paper analyzes how the factors affect the reduction,spatial transfer,and temporal transfer capabilities of the dats center loads.We would innovate the existing mechanism,model,and algorithm,and propose demand response strategies considering the data center’s load adjustment potential.The specific research content is as follows:(1)Research data center demand response strategies that consider multiple uncertainties,analyze the impact of multiple uncertainty coupling relationships on load reduction,and quantify load reduction capability.First,the power consumption models of the equipment in the data center are designed.An IT consumers-oriented load reduction capability model is established to estimate the value of load reduction capability.Then,design the utility functions of the data center and IT consumers considering four types of uncertainties,such as compensation,incentives,load reduction capability,and load reduction quantity.We use the Stackelberg game to solve the optimal solution of utility functions for the data center and IT consumers and to calculate the specific regulation plan(i.e.,the optimal incentives and load reduction quantity).Finally,take the upper and lower limits of the uncertain factors in the regulation plan,and use the interval optimization model to optimize the demand response strategies for optimistic solutions and pessimistic solutions when the data center participates in peak shaving and valley filling.Those solutions would decide the IT consumers participating in peak shaving and valley filling,the number of IT consumers,and load reduction quantity.The coupling relationship among multiple uncertain factors on the load reduction is analyzed through the interval optimization model,and the upper and lower intervals of the load reduction are derived.(2)Study data center demand response strategies under computing resource sharing mode,and analyze the limitations of computing resource ownership on the data center’s load transfer capability.First,the load of multiple data centers is divided into the day-ahead group that participates in demand response and the backup group that shares computing resources.We devise utility functions of the data center operator,the day-ahead group,and the backup group.Then,the regulation plans of the day-ahead group are formulated by the Stackelberg game theory,and the Nash bargaining game determines the computing resources shared by the backup group.Finally,the demand response strategies for the day-ahead and backup groups are optimized by using a two-stage robust optimization model.In the strategy,the day-ahead group normally participates in peak shaving and valley filling,and the backup group shares the computing resource.When a sudden workload occurs in the day-ahead group,the increased load can be transferred to the backup group through spatial migration of the workload.Based on the computing resource sharing mode,Nash bargaining game,and the two-stage robust optimization model,the influence of computing resource ownership on the load spatially transfer is solved.Meanwhile,the spatial load transfer capability is deeply explored.(3)Research data center demand response strategies that consider network reliability,and analyze the impact of low network reliability on the data center’s load spatial transfer capability.First,the traditional processing mechanism of the workload for the data center and the edge node is analyzed.We design a novel communication path selection method for network reliability,which defines the principle of disjoint communication paths among service source nodes and edge nodes,service source nodes to the data center,and edge nodes to the data center.Use a proposed exhaustive algorithm to find nodes that can guarantee network reliability(i.e.,satisfying the above principle).Finally,according to the goal that the edge nodes could cover the maximum number of service source nodes,some appropriate location for edge nodes that can guarantee network reliability is selected from the network.When the data center participates in peak shaving and valley filling,the workload is migrated from the data center to the edge node.Correspondingly,the power demand of the data center is transferred to each edge node evenly.According to the selection of edge node deployment location and the transfer model of data center load,the impact of low network reliability on the spatial load transfer capability is reduced.(4)Study data center demand response strategies that consider IT consumers’response characteristics,and analyze the impact of workload QoS adjustments on the data center’s load temporal transfer capability.Firstly,based on the QoS requirement,the data center’s workload is divided into adjustable and non-adjustable.The data center would incentivize IT consumers to adjust the QoS of workload.For instance,the workload submission and the completion times can be advanced and delayed,respectively.Then,the utility functions of IT consumers and the data center are designed,and a bi-level optimization model is proposed.At the upper level,the data center acts as a leader,releasing incentives to IT consumers while optimizing energy costs through workload scheduling and integrating renewable energy.At the lower level,the IT consumers play the follower role,analyze the incentive and QoS losses,adjust the processing time of the workload based on their response characteristics,and enlarge the processing time window of the workload.Finally,the particle swarm optimization algorithm and the mixed integer linear programming model are applied to solve the bi-level model.According to the proposed model and algorithm,improve the data center load of temporal transfer capability.Meanwhile,the ability of the data center to integrate renewable energy is heightened.