Research On Energy Management Of Data Center Microgrids Based On Stochastic Predictive Model

With the booming development of cloud computing,AI,Internet of things,5G and other information technologies,data centers,as the core infrastructure of the information industry,have entered the stage of rapid development.As the quantity and scale of data centers expand constantly,the power consumption and operating cost of data centers also increase correspondingly,and the environment is affected at the same time.In order to solve the problem of high power consumption and high cost,as well as ensure the reliability of power supply,data centers are connected to the local power grid,and are usually equipped with renewable energy generators such as wind turbines and photovoltaic,traditional generators,and energy storage system,which can form a “data center microgrid”.Considering the randomness of the renewable energy and the unique characteristics of data center workloads,the energy management of data center microgrids is studied in this paper.Firstly,this paper respectively analyzes the power consumption and power supply of the data center.By analyzing the characteristics of interactive workloads and batch workloads,the power consumption model is established.Based on the analysis of the operating characteristics and principles of all power supplies,the power supply model is established.On the basis of above analysis,the system structure of a typical data center microgrid is proposed in this paper.Combined with the basic knowledge of energy management,the energy management frame of data center microgrids is also constructed.Secondly,due to the type of the professional work,the load of data centers is more random than that of ordinary microgrids,which poses a certain challenge to the energy optimization scheduling.Therefore,this paper studies the forecasting technology of load power,and establishes a prediction model based on the long short term memory neural network algorithm.Then the actual historical load data of a certain microgrid is used to simulate on the Matlab platform,and the prediction results are compared with other prediction methods.The simulation results show that the prediction effect of the LSTM neural network model is better and the accuracy is higher than other models.Therefore,the model can be used to predict the electricity load of data centers.Finally,based on the above research and analysis,combined with the basic theory of stochastic programming and model predictive control,this paper proposes an energy optimization scheduling method of data center microgrids.Scenarios analysis method is used to describe the uncertain factors.Considering the flexible scheduling of batch workloads and the constraint conditions of all equipments in the system,the day-ahead scheduling model is established with the objective of economic optimization.In order to ensure the safety and stability of the data center connected to the power grid in the actual operating,the intra-day scheduling model is established based on the rolling optimization and feedback correction mechanism of model predictive control.The output of each equipment is adjusted on the premise of tracking the day-ahead scheduling as far as possible.To solve the model,an energy management process based on stochastic predictive model is designed in this paper.In order to explore the effectiveness of the proposed model and strategy,the simulation case of a typical data center microgrid is constructed on the Matlab platform.Through the analysis of the simulation results,it is proved that the model and strategy studied in this paper have good effects on improving the stability and economy of the data center operation.