Data Center Heat Load Prediction Based On Thermal Management Mechanis

At present,there are a lot of researches on data center electricity load,space temperature distribution in the server room,and IT equipment load flow prediction,but there are less researches on studying data center energy consumption from the perspective of thermal load.In this topic,the following studies are conducted on the composition of data center thermal load,selection of prediction methods,and establishment and verification of prediction models based on thermal management mechanisms:(1)Study of data center load characteristics:the building envelope,outdoor weather conditions,and internal equipment are important factors affecting the building heat load.In addition,data center is a building dedicated to serve electronic information technology equipment,and the host room is arranged with a large number of IT equipment,which has characteristics different from those of general public buildings:(a)IT equipment generates huge heat(sensible heat mainly)and dissipates quickly,and(b)operates dynamically and cools throughout the year.Therefore:the heat load of data center is composed of the heat load of data center building and the heat generated by IT equipment inside the data center.(2)Data center building heat load prediction research:according to the actual situation of the research object in this paper,the simulation simulation method based on Energy Plus software is adopted,based on three equations of indoor air heat balance,room internal surface heat balance and room external surface heat balance,combined with the research object to establish a model as well as boundary conditions,and five different heat transfer coefficients of external walls are selected according to the building design code to study The load variation of the data center building is studied.The simulation results show that as the heat transfer coefficient of the exterior walls decreases,the heat load and cooling load of the data center building decreases hour by hour throughout the year.(3)Data center IT equipment heat load prediction study:based on thermal management,compare data center IT equipment heat load modeling methods,where the IT equipment detailed modeling method can accurately identify the sources of IT equipment heat load generation,and use MTLAB software to build heat load for servers,network equipment,power supply and distribution equipment based on triangle link UPS load factor and UPS efficiency,and IT equipment respectively The mathematical models and the quadratic equations describing the relationship between UPS efficiency and UPS load factor are able to estimate the triangle link UPS efficiency well with a deviation of no more than 2.5%.The simulation results show that the prediction results of the server heat load model,network equipment heat load model,and supply and distribution equipment heat load model are all in the range of the percentage of IT equipment heat load in the statistical range of IT equipment heat load contributors in the data center;compared with other prediction models,the server heat load prediction model established in this paper has the maximum R-squared error value of 0.9976,the root mean square error and the average absolute percentage errors are 12.1718 and 0.02341,respectively;therefore,the heat load prediction models for IT equipment and components of data centers are reasonable,effective,and accurate.(4)Analysis of data center heat load results and energy conservation study:the appropriate increase of heat transfer coefficient of the external wall is beneficial to the reduction of data center heat load in Guiyang area;on the premise of no condensation on the inner surface of the enclosure,combined with the maximum indoor design temperature of this data center,the extreme minimum temperature of Guiyang and the design code,the heat transfer coefficient of the external wall of this data center in Guiyang area should be less than 1.492W/m2 K.The heat load of the data center is mainly generated by the heat dissipation of IT equipment:among them,servers and network equipment contribute more than 90%,so it is necessary to focus on the operation status of servers and network equipment in the data room in the data center thermal management.The detailed modeling method of IT equipment used in this paper can better predict the data center thermal load compared with traditional methods.