基于傅里叶频域增强与三重线性连接的服务器能耗预测方法

打开文本图片集

关键词：数据中心；服务器能耗；傅里叶变换；长期时序数据；三重线性连接DOI：10.15938/j. jhust. 2025.06.007中图分类号：TP391 文献标志码：A 文章编号：1007-2683（2025）06-0061-09

Abstract：Withtherapiddevelopmentof data centers，theissueofserver power consumptionis graduall highlighted.The accuracyof predictionisofgreatsignificanceforimprovingtheeficiencyofserveroperationandenergyconservationandeision reduction.This paper proposes anew server energyconsumption prediction method.This methodfirstpreprocessesand encodes the long-termtimeseriesdata，theninputsitintothehybriddecompositionmodule.Thismoduledecomposesthedataintotrendand periodic parts，simplifyingcomplextimeseriespatersandmaking themmore interpretable.Subsequently，thedecomposeddatais senttotheFourierransfoBoostingmodule（FTB）Thismoduleemphasizesthepotentialperiodicityvibrationfrequencyanddata fluctuatincharacterisicspresentintedata.Finally，llinformationisenttotriplelinaroetionmduleforprocessingBy addingtheoutputsoftelinearblocks，thefinalpredictionresultisobtainedExperimentalresultsshowthatthismetodcanectively improve the accuracy of energy consumption prediction and reduce power consumption.

Keywords：data center；serverpower consumption；fourier transformation；long-termtimeseries data；triple linear coection

0引言

随着科技的快速发展，数据中心已经成为支持现代信息社会不可或缺的角色，如在云计算、大数据分析、人工智能等领域为人们提供了必不可少的云存储、计算能力以及网络服务等，都极大地依赖于卓越和稳定的数据中心基础设施。（剩余14146字）