文档名:质子交换膜燃料电池退化预测方法
摘要:耐久度是制约质子交换膜燃料电池大规模应用的主要障碍之一,性能退化预测技术可以有效提高质子交换膜燃料电池的耐久度.该文提出一种结合小波阈值去噪方法的正则化堆叠长短期记忆网络的性能退化预测方法.通过小波阈值去噪法,获得消除噪声和尖峰后的平滑数据.针对退化数据不确定性和高度非线性导致的特征难以提取问题,引入了正则化堆叠长短期记忆网络模型,该模型通过引入参数优化算法有效地避免了过拟合风险,提高了预测精度和可靠性.为验证该方法的有效性,采用两种不同工况下的质子交换膜燃料电池老化数据进行验证.结果表明,所提方法在稳态工况下的最大误差为0.0163V,误差区间在0.5%以内;动态工况下的最大误差为0.0064V,误差区间在0.2%以内.
Abstract:Durabilityisoneofthemainobstaclestothelarge-scaleapplicationofprotonexchangemembranefuelcell(PEMFC).PerformancedegradationpredictiontechnologycaneffectivelyimprovethedurabilityofPEMFC.ThroughthestudyofPEMFCagingdata,itisfoundthattheactualPEMFCagingdataishighlynonlinear,periodicandrandom,whichmakesitdifficultforthepredictionalgorithmtoextractthefeatureseffectively.Inaddition,intheproblemofdegradationprediction,thepredictionalgorithmneedstopredictthedegradationofPEMFCunderdifferentworkingconditions,whichrequiresthepredictionalgorithmtohavestrongergeneralizationability.Tosolvetheaboveproblems,aperformancedegradationpredictionmethodofregularizationstacklongshort-termmemorycombinedwithwaveletthresholddenoisingmethod(WTD-RS-LSTM)methodisproposed.Firstly,theWTDmethodisusedtoprocesstheoriginaldata,andthesmoothdataaftereliminatingnoiseandspikesisobtainedbywaveletdecomposition,thresholdprocessinganddatareconstruction.ThentheRS-LSTMmodelisintroducedtosolvetheproblemoffeatureextractioncausedbyuncertaintyandhighnonlinearityofdegradeddata.Thegeneralizationabilityofthemodelisimprovedbyintroducingparameteroptimizationalgorithm.Themodelisstackedtoenhanceitslearningability.Forincreasethereliabilityofthemodel,Warmupstrategywasusedtodynamicallyadjustthelearningrateofthenetwork.Throughtheaboveoperations,theoverfittingphenomenonwhichmayoccurinthetrainingofthemodeliseffectivelyavoided,andthepredictionaccuracyandreliabilityofthepredictionalgorithmareimproved.Forverifytheeffectivenessoftheproposedmethod,PEMFCagingdataundertwodifferentworkingconditionsareusedforverification.Thedatasetsunderdifferentworkingconditionsaredividedintofivedifferentlengthsoftrainingsetsandtestsetstotrainandtesttheproposedalgorithm.Theverificationresultsshowthatundersteady-stateconditions,themaximumerroroftheproposedmethodis0.0163V,andtheerrorintervaliswithin0.5%.Thepredictionperformanceincreaseswiththetraininglength,andthebestpredictionperformanceisobtainedatthetraininglengthof1000h,whentheRMSEandMAPEare0.00091and0.00022,respectively.Underdynamicconditions,themaximumerroris0.0064Vandtheerrorintervaliswithin0.2%.Thebestperformancewasachievedwhenthetraininglengthwas550h,whentheRMSEandMAPEare0.00075and0.00020,respectively.Accordingtotheaboveexperimentalresultsandthecomparisonwiththeexistingtraditionalalgorithms,thefollowingconclusionsaredrawn:(1)theproposedmethodcanmakemoreaccuratePEMFCdegradationpredictionunderdifferentworkingconditionsanddifferenttraininglengths,andhasstrongergeneralizationability;(2)Comparingthepredictionaccuracyofthetwoconditionsunderdifferenttraininglengths,itisfoundthatthepredictionofPEMFCdegradationunderdynamicconditionsbytheproposedmethodisbetterthanthatundersteady-stateconditions.Therefore,theproposedmethodhasstrongerpredictionabilityunderdynamicconditions.(3)Theproposedmethodhasasimplestructure,easytodeployandissuitableforonlineapplication;(4)TheagingofPEMFCunderdynamicconditionswillproducemorerandomness,whichwillhaveagreatimpactonthestabilityofthepredictionalgorithm.
作者:汪建锋 王荣杰 林安辉 王亦春 张博 Author:WangJianfeng WangRongjie LinAnhui WangYichun ZhangBo
作者单位:集美大学轮机工程学院厦门361021集美大学轮机工程学院厦门361021;电工材料电气绝缘国家重点实验室(西安交通大学)西安710049
刊名:电工技术学报 ISTICEIPKU
Journal:TransactionsofChinaElectrotechnicalSociety
年,卷(期):2024, 39(11)
分类号:TM911TK91
关键词:质子交换膜燃料电池 性能退化预测 小波阈值去噪 长短期记忆网络
Keywords:Protonexchangemembranefuelcell(PEMFC) degradationprediction waveletthresholddenoising longshort-termmemory(LSTM)
机标分类号:TP391P208TP183
在线出版日期:2024年6月18日
基金项目:国家自然科学基金,福建省自然科学基金项目,电力设备电气绝缘国家重点实验室基金,福建省中青年教师教育科研项目质子交换膜燃料电池退化预测方法[
期刊论文] 电工技术学报--2024, 39(11)汪建锋 王荣杰 林安辉 王亦春 张博耐久度是制约质子交换膜燃料电池大规模应用的主要障碍之一,性能退化预测技术可以有效提高质子交换膜燃料电池的耐久度.该文提出一种结合小波阈值去噪方法的正则化堆叠长短期记忆网络的性能退化预测方法.通过小波阈值去...参考文献和引证文献
参考文献
引证文献
本文读者也读过
相似文献
相关博文
质子交换膜燃料电池退化预测方法 Degradation Prediction Method of Proton Exchange Membrane Fuel Cell
质子交换膜燃料电池退化预测方法.pdf
- 文件大小:
- 2.5 MB
- 下载次数:
- 60
-
高速下载
|
|