搜索
分类号密级UDC学位论文模糊神经网络在电梯群控系统中的应用研究作者姓名：张绍谦指导教师：杨卫国副教授东北大学信息科学与工程学院申请学位级别：硕士学科类别：工学学科专业名称：控制理论与控制工程论文提交日期：2009年6月30日论文答辩日期：2009年7月4日学位授予日期：2009年7月答辩委员会主席：王建辉评阅人：周玮、张军东北大学2009年6月AThesisinControlTheoryandControlEngineeringApplicationResearchofFuzzyNeuralNetworkinElevatorGroupControlSystembyZhangShaoqianSupervisor:AccociateProfessorYangWeiguoNortheasternUniversityJune2009-I-独创性声明本人声明所呈交的学位论文是在导师的指导下完成的。论文中取得的研究成果除加以标注和致谢的地方外，不包含其他人已经发表或撰写过的研究成果，也不包括本人为获得其他学位而使用过的材料。与我一同工作的同志对本研究所做的任何贡献均已在论文中作了明确的说明并表示诚挚的谢意。学位论文作者签名：日期：学位论文版权使用授权书本学位论文作者和指导教师完全了解东北大学有关保留、使用学位论文的规定：即学校有权保留并向国家有关部门或机构送交论文的复印件和磁盘，允许论文被查阅和借阅。本人同意东北大学可以将学位论文的全部或部分内容编入有关数据库进行检索、交流。作者和导师同意网上交流的时间为作者获得学位后：半年□一年□一年半□两年□学位论文作者签名：导师签名：签字日期：签字日期：东北大学硕士学位论文摘要-II-模糊神经网络在电梯群控系统中的应用研究摘要随着高层建筑物的日益增多，电梯群在高层建筑和智能大厦中的作用越来越重要，电梯群控系统已经称为国内外研究的热点。本文的目的是根据电梯群控的理论和研究焦点，应用合适的智能算法进行电梯群控制。本文首先回顾了电梯群控的发展历史和现状，介绍了电梯群控的基本理论和两种常用的电梯群控系统控制方法：模糊逻辑和神经网络，并分析了他们的优缺点。然后，本文介绍了模糊逻辑和神经网络结合的模糊神经网络，并详细分析了模糊神经网络的结构和推理过程，然后针对本文采用的模糊神经网络模型，笔者编制了相应的MATLAB程序并对程序和模型进行了仿真实验以验证其有效性和可靠性。在完成了前文的理论准备后，本文对模糊神经网络进行混合训练学习，首先使用了k-均值算法初步确定了隶属函数的中心和宽度，然后使用顺序聚类方法提取模糊规则，最后使用带有动态惯性权值的PSO对隶属函数的中心和宽度进行了优化调整。经过混合训练学习方法对模糊神经网络的训练和优化后，可以得到较为完整的模糊神经网络。此处构建的模糊神经网络模型的方法将应用于随后的电梯群控模式识别模块和派梯调度模块。按照上文提到的模糊神经网络模型构建和训练学习方法根据交通流模式识别的特点和要求建立相应的模糊神经网络，然后通过构建并训练完毕的模糊神经网络对当前交通流特征值分析以得到相应的交通流模式，从而可以根据当前交通流模式制定相应的调度控制策略。之后按照上文提到的模糊神经网络模型构建和训练学习方法根据电梯群控派梯调度系统的特点和要求建立相应的模糊神经网络，根据前面得到的交通流模式和派梯调度策略，通过模糊神经网络对当前电梯运行状态数据进行的分析从而对呼梯信号序列进行调度。最后对本文建立的电梯群控系统进行了仿真和数据验证，根据模拟实际系统的运行，验证了本文算法和模型的有效性。关键词：电梯群控系统，模糊神经网络，PSO算法，交通流模式识别，电梯群控派梯调度东北大学硕士学位论文Abstract-III-ApplicationResearchofFuzzyNeuralNetworkinElevatorGroupControlSystemAbstractWiththedevelopmentofhigh-risebuildings,elevatorgroupisplayingamoreimportantroleinhigh-risebuildingsandintelligentbuildings,sotheelevatorgroupcontrolsystem(EGCS)isnowthefocusoftheresearchersathomeandaboard.InthispapersuitableintelligentalgorithmsareappliedinthiscomplicatedsystemaccordingtothetheoryofECGSandresearchfocus.Firstlythehistoryandstatusquoarereviewed,andthebasictheoryofEGCSisintroduced,thentwocommonmethodsinEGCS:FuzzyLogicandNeuralNetwork,areputforwardandtheiradvantagesandShortcomingareanalyzed.ThenthecombineofFuzzyLogicandNeuralNetwork:FuzzyNeuralNetwork,isputforward,andalsothestructureandreasoningprocessareanalyzed.ThentheMATLABprogramismadeuseoftoverifytheeffectivenessandreliabilityoftheFuzzyNeuralNetworkmodelintroducedinthispaper.Afterthepreparationoftheory,amixedlearnandtrainmethodisusedintheFuzzyNeuralNetworkmodel.Firstlyk-meansalgorithmisusedtogetthepreliminarycentersandwidthsofthemembershipfunctions,thenthefuzzyrulesareabstractedbythemeansofsequenceclustering,atlast,PSOwithdynamicinertiaweightisappliedintheoptimizationandadjustmentofcentersandwidthsofthemembershipfunctions.Whenthemixedlearnandtrainmethodisfinished,aFuzzyNeuralNetworkmodelisbuilt.ThemeansusedheretobuildandtrainaFuzzyNeuralNetworkwillbeappliedinthefollowingpatternrecognitionoftrafficmoduleandelevatorschedulingmodule.AtfirstthismeansisusedtobuildaFNNforthepatternrecognitionoftrafficaccordingtotherequestandfeatureofthismodule,thenaresultofpatternrecognitionoftrafficthroughanalysingthefeaturedataoftrafficisgot,sothecorrespondingstrategywillbedesighedaccordingtothefeatureoftraffic.ThethesamemeansisusedtobuiltaFNNfortheelevatorschedulingmodule,accordingtothepatternrecognitionoftrafficandthecorrespondingstrategy,withthehelpofFNN,theelevatorschedulingplanisgotthroughanalyzingthestatedataofelevatorsAtlast,theEGCSissimulatedandverifiedwithexperimentaldata,theresultprovestheeffectivenessofthealgorithmandmodelputforwardandbuiltinthispaper.Keywords:elevatorgroupcontrolsystem,FuzzyNeuralNetwork,PSOalgorithm,thepatternrecognitionoftraffic,elevatorscheduling东北大学硕士学位论文目录-IV-目录声明······················································································································Ⅰ摘要······················································································································IIAbstract······················································································································III第1章绪论························································································································11.1电梯群控系统的发展和现状······························································································11.2电梯群控系统概述··············································································································11.2.1电梯群控系统的特征···································································································11.2.1电梯群控系统的性能评价指标···················································································31.2.1乘客交通流数据获取方法···························································································41.3本文的目标和研究重点······································································································41.4本章小结··························································································································