一种EKF-WLS-SVR与混沌时间序列分析的瓦斯动态预测新方法.pdfVIP

第28卷 第 1期 传 感 技 术 学 报 Vo1．28 No．1 2015年 1月 CHINESEJOURNALOFSENSORSANDACTUATORS Jan．2015 A New M ethodofM ineGasDynamicPredictionBasedon EKF．W LS．SVR andChaoticTimeSeriesAnalysis FU Hua ，ZIHal，MENG Xiangyun2 SUN L1 ， (1．FacultyofElectricalandControlEngineering，LiaoningTechnicalUniversity，HuludaoLiaoning125105，China； 2．JiazhouPowerSupplyCompanyofLiaoningElectricalPowerCompanyofStateGrid，JinzhouLiaoaing121000，China) Abstract：Consideringhighlychaoticcharacteristicofgasconcentration time series，the differentialentropyratio methodwasadoptedtosynchronouslydetermineoptimalembeddeddimension anddelaytimeSOastorestorethe system statespaceofgasemission．Thesampleswithnoiseeliminationcamefrom thegasconcentrationserieswhich wascollectedbywirelesssensornetworks．Extendedkalmanfilteralgorithm(EKF)withtuningfactorwasproposed torapidlyoptimizetheregularizationparameterYandthenuclearparameterf，oftheweightedleastsquaressupport vectorregression(WLS-SVR)．PeriodicallyupdatedtrainingsampleswereusedtoestablishtheEKF—WLS—SVR couplingalgorithm—baseddynamicpredictionmodeltoaccuratelypredictgasconcentrationinthefuture．Thesimu- lationoftheMATLAB showsthatEKFisgreatlyhelpfulforimprovingWLS-·SVR fittingprecisionandlearningeffi-- ciency．Thecouplingmodelispracticallyusefulandoutperformsotherpredictionmodelsintemr sofpredictionac— curacyandrobustness． Keywords：dynamicprediction；gasconcentration；chaoticcharacteristic；wirelesssensornetworks；WLS—SVR；EKF EEACC：7230 doi：10．3969／j．issn．1004-1699．2015．01．022 一 种EKF．WLS．SVR与混沌时间序列分析的瓦斯动态预测新方法术 付 华 ，訾 海 ，孟祥云 ，孙 璐 (1．辽宁工程技术大学电气与控制工程学院，辽宁葫芦岛125105；2．国网辽宁省电力有限公司锦州供电公司，辽宁锦州 121000) 摘 要 ：针对瓦斯浓度时间序列高度的混沌特性，采用微熵率法同步确定最优的嵌入维数与延迟时间，还原瓦斯涌出系统状 态空间。以无线传感网络系统采集并经降噪处理后的瓦斯浓度序列作为样本。提Hj利用带有整定因子的扩展卡尔曼滤波器 (E