简介:摘要:电池管理系统(BMS)可以延长电池寿命,但它取决于所采用方案的准确性。已经开发了不同的技术来通过监控电池的健康状态(SOH)来增强BMS。本文采用循环计数法对电池电压的检测进行了分析,并与人工神经网络这种启发式方法进行了比较。所提出的人工神经网络方法的优点是可以在不将电池与负载断开的情况下监测SOH。此外,人工神经网络的采样数据来自各种技术,包括开路电压(OCV)法、环境温度测量和谷点检测。采用前馈反向传播算法来达到实时监控实验室的目的。结果表明,前馈神经网络(FFNN)在用更多的采样数据训练时,可以获得对SOH的精确估计。
简介:将分支前馈神经网络(BFNN)运用于数字字符的模式识别问题中,其某些性能优于标准反向传播(BP)网络。BFNN的隐层神经元与输出神经元之间为分组对应关系,采用的学习算法与标准BP算法类似。BFNN可以根据样本的可分性构建最适宜的网络结构。在对大规模、分类复杂的样本进行识别时,性能优于标准BP网络。
简介:Dynamicnodecreationandfastlearningalgorithmforahybridfeedforwardneuralnetwork.Flight-pathanglecontrolvianeuro-adaptiveBackstepping.Locallearningframeworkforhandwrittencharacterrecognition.Maximizingmarginsofmultilayerneuralnetworks.ModularnetworkSOMself-orgmlizingmapofasystemsgroupinfunctionspace.
简介:ApplicationoftheRTNNmodelforasystemidentification,predictionandcontrol;AssociativeMemoryUsingRatioRuleforMulti-valuedPatternAssociation;Batch-to-BatchModel-basedIterativeOptimisationControlforaBatchPolymerisationReactor;BehaviouralPlasticityinAutonomousAgents:AComparisonbetweenTwoTypesofController;ChannelEqualizationUsingComplex-ValuedRecurrentNeuralNetworks;Classificationofnaturallanguagesentencesusingneuralnetworks;Combiningarecurrentneuralnetworkandtheoutputregulationtheoryfornon-linearadaptivecontrol。
简介:ConfigurablemultilayerCNN-UMemulatoronFPGA;Cortically-inspiredVisualProcessingwithaFourLayerCellularNeuralNetwork;Effectofcouplingresistorsonsteadypatternsincoupledoscillatornetworks;Exponentialconvergenceestimatesforneuralnetworkswithmultipledelays;FEATUREEXTRACTIONINEPILEPSYUSINGACELLULARNEURALNETWORKBASEDDEVICEFIRSTRESULTS;FurtherResultsontheStabilityofDelayedCellularNeuralNetworks;Globalstabilityanalysisindelayedcellularneuralnetworks;ImageedgedetectionusingadaptivemorphologyMeyerWavelet-CNN。
简介:PredictionoftheDimensionalChangesduringSinteringusingBackpropagationAlgorithm,Predictionofthenextstockpriceusingneuralnetwork-extractionthefeaturetopredictnextstockpricebyfiltering,Pulsemodeneuronwithpiecewiselinearactivationfunction,Remarksonmultilayerneuralnetworksinvolvingchaosneurons……
简介:AnewapproachtogenerateAself-organizingfuzzyneuralnetworkmodel.Anonlinearcombiningforecastmethodbasedonfuzzyneuralnetwork.Anovelclustermethodinfrizzyneuralnetworks.AnovelrobustPIDcontrollerdesignbyfuzzyneuralnetwork.Arecurrentfuzzyneuralnetwork:learningandapplication.Astudyofchatterpredictioninendmillingprocess(fuzzyneuralnetworkmodelwithinputsofcuttingconditionsandsound.Aweightedfuzzyreasoninganditscorrespondingneuralnetwork.