学科分类
/ 1
6 个结果
  • 简介:Inthecontextoftowermeasuredradiationdatasets.followingthecorrectionprinciplemeetingadiagnosticequationindataqualitycontrolandintermsofatechniqueformodelconstructionondataandANN(artificialneuralnetwork)retrievalforBPcorrectionofradiationmeasurementswithrougherrorsavailable,aBPmodelispresented.Evidencesuggeststhatthedevelopedmodelworkswellandissuperiortoaconvenientmultivariatelinearregressionmodel,indicatingitswideapplications.

  • 标签: artificial neural network(ANN) BP(backward propagation)net ROUGH
  • 简介:ThispaperexplorestheapplicationofArtificialIntelligent(AI)techniquesforclimateforecast.ItpresentsastudyonmodellingthemonsoonprecipitationforecastbymeansofArtificialNeuralNetworks(ANNs).UsingthehistoricaldataofthetotalamountofsummerrainfallovertheDeltaAreaofYangtzeRiverinChina,threeANNsmodelshavebeendevelopedtoforecastthemonsoonprecipitationinthecorrespondingareaoneyear,five-year,andten-yearforwardrespectively.Performancesofthemodelshavebeenvalidatedusinga'new'datasetthathasnotbeenexposedtothemodelsduringtheprocessesofmodeldevelopmentandtest.Theexperimentresultsarepromising,indicatingthattheproposedANNsmodelshavegoodqualityintermsoftheaccuracy,stabilityandgeneralisationability.

  • 标签:
  • 简介:一个人工的神经网络(ANN)被用来用鸣温度侧面的很多时间/军刀的手足为中间的空气建模。一个三层的前馈控制网络基于背繁殖(BP)被选算法。当温度是产量向量时,纬度,经度,和高度被选择为网络的输入向量。温度观察在从通过2007年3月16日,它在一样的卫星的1月13日的时期期间偏航,作为样品被拿训练ANN。结果建议网络为为温度的空间变化建模有高质量。在从流行实验NRLMSISE-00模型的ANN产量和那些之间的量的比较说明他们的通常一致的特征和一些特定的差别。NRLMSISE-00模型鈥檚带的吝啬的温度由锝太高?行星的波浪的stratopause附近的K-10K,和振幅和阶段第一号的活动在从在45鈥?上面的ANN模拟的一些方面是不同的0km,建议改进为更多的精确模拟在NRLMSISE-00模型被需要在附近并且在stratopause上面。关键词人工的神经网络-中间的空气-建模-背繁殖算法-NRLMSISE-00模型

  • 标签: 人工神经网络 大气层 建模 温度分布 反向传播 前馈网络
  • 简介:TwofieldexperimentswereperformedinordertodissipatethefogatWuqingDistrictofTianjininNovemberandDecemberof2009.Hygroscopicparticleswereseededtodissipatefogdropletson6-7November,2009.Liquidnitrogen(LN)wasseededintothenaturalsupercooledfogintheexperimentsof30November–1December,2009.Significantresponsewasfoundafterseeding.Significantchangeswereobservedinthemicrostructureoffoginthefieldexperiments.Theoffogdropletchangeddramatically;itincreasedfirstandthendecreasedafterseeding.RemarkablevariationalsowasfoundintheLiquidWaterContent(LWC)andinthesizeoffogdroplet.TheDropletSizeDistribution(DSD)offogbroadenedduringtheseedingexperiments.TheDSDbecamenarrowaftertheseedingended.Afterseeding,thedropletswerefoundtobeatdifferentstagesofgrowth,resultinginatransformofDSDbetweenunimodaldistributionandbimodaldistribution.TheDSDwasunimodalbeforeseedingandthenbimodalduringtheseedingexperiment.Finally,theDSDbecameunimodallydistributedonceagain.

  • 标签: artificial DISSIPATION of FOG microphysical changes
  • 简介:A2m~3isothermalcloudchambermainlyforicenucleationresearchisdescribedinthispaper.Itsstructure,attachedinstrumentsandexperimentalproceduresarealsopresented.TheexperimentsofdeterminingtheicenucleieffectivenessfortheAgl-containingaerosolsproducedbythreeformulationshavebeenconductedandtheresultshavebeencomparedwiththoseoftheCSU960-literisothermalcloudchamber.Allexperimentalresultsshowthatthechamberhasadvantagesofstableperformanceandreproducibility.Itwouldbeexpectedtobecomeausefulexperimentalfacilityforicenucleationresearch.

  • 标签: NUCLEATION ISOTHERMAL instruments cloud ATTACHED AEROSOLS
  • 简介:从卫星导出的大气的不稳定性信息在预报的短期的天气起一个重要作用,特别预报严重对流暴风雨。为为朝鲜多种用途的地球同步的卫星节目的天气卫星的下一代,一台新成像仪器被开发了。尽管这台成像仪器没被设计执行完整的发出声音的使命,它的能力是有限的,它的多光谱的红外线的隧道在垂直发出声音上提供信息。为了从大部分充分利用观察数据,改进了成像器的空间与时间的分辨率,为大气的不稳定性的推导的一条人工的神经网络途径的可行性被调查。有前馈控制和背繁殖训练算法的多层的视感控器模型显示出相当敏感回答到训练数据集和模型建筑学的选择。通过有7197独立侧面的小心地选择的训练数据集的广泛的性能测试,模型体系结构被选择分别地为隐藏的节点,时代的数字,和学习的率的数字是12,5000,和0.3。选择模型给330Jkg的一个吝啬的绝对错误,RMSE,和关联系数-1,420Jkg-1,并且0.9分别地。可行性进一步从一台类似的仪器经由模型的申请被表明到真实观察数据与计划成像器有可比较的观察隧道。

  • 标签: 下线 服务 迁移