Page 51 - 2023年第54卷第3期
P. 51
[18] 刘媛媛,刘业森,郑敬伟,等.BP神经网络和数值模型相结合的城市内涝预测方法研究[J].水利学报,
2022,53(3):284 - 295.
[19] 刘鲭洁,陈桂明,刘小方,等.BP神经网络权重和阈值初始化方法研究[J].西南师范大学学报(自然科
学版),2010,35(6):137 - 141.
[20] 张忠波,何晓燕,耿思敏,等.改进的粒子群算法在水库优化调度中应用[J].中国水利水电科学研究院
学报,2017,15(5):338 - 345.
[21] 陈博文,邹海.总结性自适应变异的粒子群算法[J].计算机工程与应用,2022,58(8):67 - 75.
[22] 仝秋娟,赵岂,李萌.基于自适应动态改变的粒子群优化算法[J].微电子学与计算机,2019,36(2):6 -
10,15.
[23] 黄璇,郭立红,李姜,等.改进粒子群优化 BP神经网络的目标威胁估计 [J].吉林大学学 报 (工 学 版),
2017,47(3):996 - 1002.
[24] 韩信,张宝忠,魏征,等.考虑气象因子不确定性的参考作物蒸散量预报方法[J].中国水利水电科学研
究院学报,2021,19(1):33 - 44.
[25] 王维博,林川,郑永康.粒子群算法中参数的实验与分析 [J].西华大学学报 (自 然 科 学 版),2008(1):
76 - 80,105 - 106.
[26] 龙远,邓小龙,杨希祥,等.基于 PSO - BP神经网络的平流层风场短期快速预测[J].北京航空航天大学
学报,2022,48(10):1970 - 1978.
[27] 高昶霖,宋燕利,左洪洲,等.基于动态权重的自适应 PSO - BP神经网络焊接缺陷成因诊断[J].焊接学
报,2022,43(1):98 - 106.
[28] COURTICEG,BAUERB,CAHILLC,etal.Acategoricalassessmentofdose - responsedynamicsformanaging
suspendedsedimenteffectsonsalmonids[J].ScienceoftheTotalEnvironment,2020(807):150844.
[29] XUF,BAOLIGAOB,CHENX,etal.Shortinformativetitle:Quantitativeassessmentofacuteimpactsofsus
pendedsedimentoncarpintheYellowRiver[J].RiverResearchandApplications,2018,34:1298 - 1303.
[30] TRITTHARTM,HAIMANNM,HABERSACKH,etal.Spatio - temporalvariabilityofsuspendedsedimentsinrivers
andecologicalimplicationsofreservoirflushingoperations[J].RiverResearchandApplications,2019,35:918 - 931.
Apredictionmethodfortheimpactofhyper - concentratedflowonfishesbased
ontheIPSO- BPneuralnetwork
1
1
1
1
2
1
LIXiaochen,Baiyinbaoligao,LIXiangdong,XUFengran,MUXiangpeng,DONGZhiqiang
(1.StateKeyLaboratoryofSimulationandRegulationofWaterCycleinRiverBasin,
ChinaInstituteofWaterResourcesandHydropowerResearch,Beijing 100038,China;
2.QinghaiInstituteofWaterConservancyandHydro - electricpowerDesignCo.,Ltd,Xining 810001,China)
Abstract:Thehyper - concentratedflowprocessmaycausenegativeimpactsonfishesandotheraquaticanimals
duringreservoirsedimentflushing.Nevertheless ,thereisalackofstudiesoncorrespondingquantitativeassessment
methodsforthedegreeofimpactofthehyper - concentratedflow.Inordertopredictandevaluatetheimpactofres
ervoirsedimentdischargeprocessesondownstream fish , thispaperusesexperimentaldatafrom thestudyof
survivalcharacteristicsofYellowRiver GymnocyprisEckloniandCyprinusCarpioinhyper - concentratedflowand
establishedanIPSO - BPneuralnetwork - basedmethodforpredictingtheimpactonfishmortality ,andtakesinto
accounttheeffectsofsuspendedsedimentconcentration,medianparticlesize,dissolvedoxygen,exposuretime,
watertemperature,andotherfactorsonfishsurvival.Thepredictionerrorofthetargetfishmortalityislessthan
6%.Inthispaper,theIPSOalgorithm,whichiscloselycoupledwithBPneuralnetworkandintroducesdynamic
parametersandvariationalperturbations ,hasbetterpredictionabilitythanBPandPSO - BPneuralnetworks,and
itsaccuracyissignificantlyimprovedwhencomparedtoexistingStressIndex(SI),SeverityofIllEffect(SEV),
andmultivariatefittingevaluationmethodsathomeandabroad.Theanalysisdemonstratesthatthepredictionmeth
odproposedinthispapercanaccountforthesituationinwhichfishmortalityinhyper - concentratedflowisgov
ernedbyacombinationofmultipleenvironmentalfactorswithcomplexcorrelationsamongmultiplefactors.Thispa
perprovidesanewmethodforassessingtheimpactofhyper - concentratedflowonfishes.
Keywords:IPSO - BPneuralnetwork;hyper - concentratedflow;fish;mortality;predictionmethod
(责任编辑:韩 昆)
— 3 0 1 —