Thermal Environment Simulation And Optimization Control Of Row-level Air Supply Data Center

With the rapid development of the data center industry,the number and scale of data centers are growing rapidly.In order to save the floor space of the data center,the layout of IT equipment in the data center is also becoming more and more compact,which makes some high-heat-generating equipment form local hot spots,causing the equipment to operate in a harsh environment,which affects the reliability of the data center.Since the distribution of local hotspots in the computer room changes with the use of IT equipment and the air-conditioning air supply,how to eliminate these hotspots and achieve energy saving has become a new challenge for data center air-conditioning system control.In order to solve this problem,this thesis firstly conducts research on data center thermal environment simulation and temperature field rapid prediction methods.Then,based on the rapid temperature field prediction,this thesis studies the optimization control strategy of the data center air conditioning system.In this thesis,a cloud computer room in a university is used as the research object,and a three-dimensional model and a computational fluid dynamics model are established.The temperature field information in the data center is obtained by numerical simulation.In order to verify the accuracy of the numerical simulation model,a mobile temperature measurement and data acquisition test bench is designed and built,and the actual data is visualized based on the spatial interpolation algorithm.By comparing the measured temperature with the simulated value,it is found that the relative error between the measured value and the simulated value of 86.4% of the measured points is within ± 20%,which verifies the accuracy of the numerical simulation model.On the basis of numerical simulation,the proper orthogonal decomposition method is adopted in this thesis.The training data sets to train the eigen-orthogonal decomposition model is provided by the numerical simulation model,and finally a rapid prediction model of the inlet and outlet air temperature of the cabinet in the data center is established.This thesis analyzes the error of the rapid prediction model by setting up four test conditions and comparing the simulated value of the numerical calculation with the predicted value.The comparison results show that the relative error of 85.5% of the predicted value is within ± 10%,and that of 96.5% of the predicted value is within ±15%,indicating that the rapid prediction model can meet the control needs.This thesis proposes an optimal control strategy for the supply air temperature of the data center air conditioning system.The idea of simulating annealing optimization algorithm is combined with the fast prediction model,and the Monte Carlo criterion is introduced to optimize the supply air temperature.Finally,the numerical simulation model was used to verify the optimized control strategy.The verification results show that under the conventional control strategy,the maximum temperature of the hot spot with the high constant supply air temperature is 33.56 ?,which is higher than that is 32.19? under the optimized control strategy.That indicates that the optimization control strategy can optimize the supply air temperature according to the change of working conditions in real time to meet the server's heat dissipation requirements.Under the condition of low constant supply air temperature,the maximum temperature value of the hot spot is 29.67 ?,which is lower than the value under the optimized control strategy.This is because its supply air temperature is lower than the optimized control strategy,which proves the energy-saving effect of the optimized control strategy.The optimized control strategy proposed in this thesis can achieve the purpose of energy saving while satisfying the cooling requirements of data center racks.