搜索
天津理工大学中环信息学院2011届本科毕业设计说明书I基于神经网络的PID控制器设计摘要本文以提高控制器的控制效果为目标，将神经网络与PID控制相结合，分别对单变量系统和多变量系统的神经网络结构的PID控制进行了深入研究和探索。对单变量系统，将预测控制的思想和神经网络PID控制的思想结合起来，用多步预测性能指标函数去训练神经网络把纬d器的权值，就构成了多步预测性能指标函数下的神经网络PID控制系统:本文对此系统进行了改进，引入了新的多步预测性能指标函数，同时对神经网络辨识器部分，采用了更适合于实时控制的动态递归神经网络来代替原方案中的多层前向网络对其辨识。仿真结果表明此改进方法比原方法及带辨识器的神经网络PID控制方法具有更好的响应性能。对多变量系统，首先研究了基于多步预测性能指标函数下多变量系统的神经网络PID控制，并给出仿真实例及结论;接着研究了PID神经元网络多变量控制器的结构和计算方法:它是由并列的多个子网络组成，当控制系统有n个被控变量，子网络就有n个。每个子网络的输入层接受系统的给定信号和对象的输出信号;隐含层由比例元、积分元、微分元组成，实现PID运算;输出层实现规律的综合:为加快权值调整速度对输出层权值采用最小二乘法进行调整，代替原方案中的梯度法，仿真结果表明此系统具有良好的自学习和自适应解祸性能。关键词:神经元网络PID控制多步预测性能指标函数动态递归神经网络单变量系统多变量系统ThePIDControllerWasBasedOnNeuralNetworkDesign天津理工大学中环信息学院2011届本科毕业设计说明书IIABSTRACTInordertoenhancetheperformanceofthecontroller,thispapercombinedtheneuralnetworkandPIDcontrol,anddeeplystudiedtheneuralnetworkPmcontrollerbasedonsingle-variableandmufti-variablesystem.Forsingle-variablesystem,theneuralnetworkPIDcontrollerbasedmufti-steppredictiveperformancetargetfunctioncombinedthepredictivecontrolideaandtheneuralnetworkPIDcontrolidea.Itusesmulti-steppredictiveperformancetargetfunctiontotraintheweightsofneuralnetworkPIDcontroller,Thispaperimprovedthesystem:Itusesnewmufti-steppredictiveperformancetargetfunctiontotraintheweights,anditusesthedynamicrecursionneuralnetworkinsteadofmultiplayerfeedforwardneuralnetworkthatisfurthermorefitforreal-timecontroltoidentifythepartofneuralnetworkidentification.ThesimulatingresultsshowsthatthismethodhasbetterresponseperformancethantheneuralnetworkPIDcontrolmethodwithidentificationFormufti-variablesystems,atfirst,ThispaperstudiedtheneuralnetworksPIDcontrollerbasedmufti-variablesystemsusingmufti-steppredictiveperformancetargetfunction,Afterstudyingthesystem'ssimulatinginstances,Igottheresults;ThenthispaperstudiedthestructureandarithmeticofthePIDneuralnetworkmultivariablecontroller.Itismadeupofparatacticmufti-sub-network,iftherearencontrolledvariablesincontrollingsystem,thesub-networksthenwillhaventoo.Theinputlayerofeachsub-networkacceptedthepresentsignalofthesystemandtheoutputsignalofcontrolledobject;Thehiddenlayerthatismadeupofproportion,integralanddifferentialthreepartsrealizesPIDoperation;Theoutputlayerrealizestheintegrationoftherules;anditsoutputlayer'sweightswereadjustedusingtheleastmeansquaresinsteadofgradsarithmeticinordertoquickentheregulativespeedoftheweights,theresultsshowthatthesystemhasmuchhigherperformanceofself-studyingandself-adapting.Keywords:NeuralnetworkPIDControlMulti-steppredictiveperformancetargetfunctionDynamicrecursionneuralnetworkSingle-variablesystemMultivariablesystem天津理工大学中环信息学院2011届本科毕业设计说明书III目录第一章引言.................................................................11.1课题背景及研究意义.....................................................11.2课题当今的研究现状.....................................................11.3本文的结构组成.........................................................2第二章PID控制器的基本原理...............................................32.1PID控制器..............................................................32.1.1PID原理...........................................................32.1.2PID各参数的作用...................................................42.2数字PID控制...........................................................42.2.1控制器的组成......................................................52.2.2典型的PID控制器..................................................52.3PID参数整定............................................................52.4小结...................................................................6第三章神经网络的基本原理.................................................63.1神经网络的模型结构.....................................................73.2几种典型的学习规则.....................................................83.2.1无监督的Hebb学习规则.............................................93.2.2有监督的Delta学习规则............................................93.3几种典型的神经网络.....................................................93.3.1BP神经网络........................................................93.3.2RBF神经网络.....................................................123.3.3CMAC神经网络....................................................143.4小结..................................................................16第四章神经网络PID控制基本原理以及应用................................164.1基于BP神经网络的PID控制.............................................174.1.1BP神经网络整定原理...............................................174.1.2MATLAB的背景和发展...............................................214.1.3MATLAB的工作环境.................................................224.1.4常规PID控制系统.................................................234.1.5基于BP神经网络的PID控制系统....................................244.2RBF神经网络和CMAC神经网络PID控制....................................304.2.1RBF神经网络PID控制..............................................304.2.2CMAC神经网络PID控制.............................................314.3小结...................................................................31天津理工大学中环信息学院2011届本科毕业设计说明书IV第五章绪论................................................................32参考文献.....................................................