Research On Optimal Dispatch Of Data Center Microgrid Based On Distributed-stochastic Model Predictive Control

Tracking changes in electricity prices and generating rational decisions on electricity use to achieve power supply reliability and cost-effective management is a major method for optimal dispatching of new power systems.With the rapid development of informatization,more and more data centers exist in the form of microgrid clusters.However,the penetration rate of renewable energy has increased,and the intermittent power generation characteristics are difficult to meet the extremely high power supply reliability requirements of data center microgrids.At the same time,the uncertain factors generated by various random operation scenarios will also produce unpredictable random deviations in the source network load prediction.In addition,the high energy consumption and electricity consumption characteristics of data centers increase the level of carbon emissions,which is not conducive to the construction of a clean and sustainable new power system.Therefore,using the multi-time-step rolling optimization idea,real-time feedback eliminates random errors in the system,and realizes windsolar storage and load combined production.Reasonable allocation of resources is the key to optimal scheduling of data center microgrids with high reliability,economy and flexible response.The research content of this paper is mainly divided into the following aspects:(1)The structural composition and internal energy consumption of the data center are analyzed;based on the power consumption-power supply characteristics of the data center,the workload of the data center is divided according to the different power consumption characteristics and task processing mechanisms,and a scheduler is established.Optimize the mathematical model;build a data center microgrid collaborative energy supply framework according to the power generation characteristics of various typical power sources in the microgrid;finally,establish a data center power supply optimization model based on the interconnected microgrid under the combined production of wind and solar storage.(2)A multi-time-step optimization scheduling strategy for microgrids based on Model Predictive Control(MPC)is proposed.In the day-ahead scheduling stage,based on the timeof-use electricity price,the “peak-shaving and valley-filling” characteristics of energy storage,and the forecast results of the output of renewable energy power generation,a day-ahead scheduling model is established with the lowest system operating cost,and the optimal unit output plan and energy storage Charge and discharge management;in the intraday scheduling stage,the MPC-based rolling optimization method is used to feedback and correct the planned value for the day before,which realizes the accurate and fast tracking of the planned value curve for the day before.Finally,in the microgrid model of wind-solar-load cogeneration in MATLAB,the scheduling economy,prediction accuracy,fast response and practical applicability of the proposed multi-time-step optimization strategy are simulated and verified,and different algorithms are compared and analyzed.computing performance.(3)An optimal scheduling strategy for data center microgrid clusters considering uncertainty is proposed.Firstly,the randomness of renewable energy is probabilistically processed,and the uncertainty risk index is established,which is included in the optimization objective function,and the data center day-ahead stochastic model predictive control(SMPC)optimization scheduling model is established;secondly,considering the distribution This paper proposes a multi-agent-based distributed optimization scheduling algorithm for data center microgrid clusters,and constructs a data center microgrid cluster based on DistributedStochastic Model Predictive Control(D-SMPC).Two-layer optimal scheduling architecture;finally,considering the uncertainty factors caused by the possibility of running various scenarios in the microgrid,a large number of random scenarios are generated in MATLAB,and based on the multi-scenario reduction technology,the proposed scenario is verified in the final reduced scenario set.Robustness of optimization algorithms.