Fast Reconstruction and Prediction Method of Three-dimensional Temperature Field in Data Center Room Based on Transformer
The rationality of airflow organization inside the data center is related to the energy consumption of the cooling system,the performance of IT equipment,and the safety of the thermal environment.Traditional CFD simulation is generally at the hourly level,with limited usage scenarios,making it difficult to meet the timeliness requirements of real-time operation for temperature field prediction.It presents a fast reconstruction and prediction method for the three-dimensional temperature field of computer rooms based on Transformer.By integrating deep learning with traditional CFD,the prediction time can be reduced to the level of minutes and seconds,and the global average prediction error can be controlled within 5%,which enables CFD simulation to be effectively applied from the design stage to the operation and maintenance stage,supporting intelligent operation services.
Data centerTemperature distribution predictionTransformerMachine learningCFD
万方数据
维普
