Research On Tie-line Power Smoothing Algorithm For Multiple Distributed Resource Controllable Data Centers

With the rapid development of information technology,the scale of Internet data has exploded,various data centers have spread across the globe.The proliferation of data has made the capacity of a single data center seem relatively weak.Enterprises have begun to collect and process data by establishing and connecting multiple data centers.Geo-distributed data centers have gradually become an important platform to support large-scale data analysis and application services.As the number of data centers increses,the huge energy consumption and carbon emissions of data centers are intensifying.Renewable energies,such as wind and photovoltaic power,has become an effective alternative to fossile energy,as a result of their wide distribution range,high utilization and low pollution.Many IT organizations and cloud service providers have been planning or beginning to introduce renewable energies into data centers.However,the randomness of the computing load and the volatility of the renewable energies in geo-distributed data centers are often superimposed on each other and transmitted to the tie-line,through which the power grid supplies electricity to data centers,causing drastic power fluctuations.Not only the tie-line power fluctuations pose a threat to the normal operation of various facilities in data centers,but also impact the frequency stability of power grid.From the perspective of power generation and utilization,a data center and its connected power grid,renewable energy system,energy storage system can be modeled together as an autonomous intelligent microgrid.On the other hand,considering computing and communication,geographically distributed data centers can be regarded as a resource load balancing system connected by the communication network.This paper first studies computing tasks,UPS batteries and other loads within a green data center and establishes the energy consumption model of the data center micro-grid.Then the information exchange and data transmission characteristics between geodistributed data centers are analyzed,a geographically distributed data center interconnection model is build.Finally,two kinds of loads,computational loads that can be migrated across space and battery packs with dynamically adjustable power,are proposed as controllable resources.Based on the work above,a tie-line power optimization method for multiple data centers is proposed,which uses a two-stage Butterworth low-pass filter to develop two-level target tie-line power of data centers.Task migration and delay is performed between data centers to response to the firstlevel target tie-line power,UPS charge and discharge power is controlled to realize the seconde-level target tie-line power.With the two-stage optimization,tie-line power fluctuations of geo-distributed data centers can be effectively smoothed.The experimental part in this paper has verified the optimization effect of the proposed two-stage strategy on the tie-line power of multiple data centers.The basic experiment is carried out by taking four data centers in different regions as examples.Results show that tie-line power fluctuations of each data center are well suppressed by implementing the method proposed above.In the three sets of comparative experiments of different renewable energy penetration,different initial resource utilization,and different proportion of delay-tolerate tasks,it is verified that the method remains valid in more complex scenarios.