基于改进的PSO和模糊RBF神经网络的MBR膜污染预测

2018年软件2018,V〇1.39,No.8第39卷第8期_________________________COMPUTERENGINEERING&SOFTWARE________________________国际IT传媒品牌I金项玛办文金项玛办文基于改进的PSO和模糊RBF神经网络的MBR膜污染预测陶颖新，李春青，苏华陶颖新，李春青，苏华(天津工业大学计算机科学与软件学院）摘要：为了提高对MBR膜通量的预测精度，采用模糊径向基函数（RBF)神经网絡建立网絡预测模型，并釆用改进的粒子群（PSO)算法进行优化。采用模糊推理过程与RBF神经网络所具有的函数等价性，统一系统函数。在利用改进的PSO算法对模糊RBF神经网絡进行训练时，先利用改进PSO算法得到模糊RBF神经网絡的初始权值和阈值，然后对其进行二次优化得到最终的权值和阈值。实验仿真结果表明：本文的这种方法，缩短了响应时间，稳态误差很小，能够与膜通量的期望值更好的拟合，更好的预测膜通量。关键词：MBR;PSO;RBF中图分类号：TP39文献标识码：ADOI:10.3969/j.issn.l003-6970.2018.08.012本文著录格式：陶颖新，李春青，苏华.基于改进的PSO和模糊RBF神经网络的MBR膜污染预测[J].软件，2018,39(8):52-56PredictionofMBRMembranePollutionBasedonImprovedPSOandFuzzyRBFNeuralNetworkTAOYing-xin,LIChun-qing,SUHua{CollegeofComputerScienceandSoftware,TianjinPolytechnicUniversity,Tianjin,China)【Abstr狀t】：InordertoimprovethepredictionaccuracyofMBRmembraneflux,usingafUzzyRadialBasisFunc_tionneuralnetworktoestablishanetworkpredictionmodel,andusetheimprovedParticleSwarmOptimization(PSO)algorithmtooptimize.ThefunctionalequivalenceofthefuzzyinferenceprocessandtheRBFneuralnetworkisusedtounifythesystemfunction.WhenusingamodifiedPSOalgorithmtotrainafuzzyRBFneuralnetwork,First,usingtheimprovedPSOalgorithmtoobtaintheinitialweightsandthresholdsofthefuzzyRBFneuralnet­work,andthenperformasecondoptimizationonthemtogetthefinalweightsandthresholds.Theexperimentalsimulationresultsshowthatthismethodofthispapershortenstheresponsetime，hasasmallsteady-stateerror,andcanbetterfittheexpectedvalueofthemembranefluxandbetterpredictthemembraneflux.【Keywords】：MBR;PSO;RBF0IntroductionThemembranebioreactor(MBR)isanewwas­tewatertreatmenttechnology,whichcombinesmem­braneseparationtechnologywithbioreactortechnol-ogy.MembranefluxisanimportantparameterintheMBRstudy.Membranefluxreflectsmembranefouling.ThepredictionofmembranepollutionthroughtheestablishmentofpredictionmodelshasbecomeanimportantresearchdirectioninMBRsimulation.Mostofthecommonlyusedpredictionmodelshavesomedefects,suchasinsufficientanalysisofthemembranefoulingmechanismandpoorpredictionaccuracy.I