Page 75 - 《水资源与水工程学报》2024年第2期
P. 75
!35 "!2 # & ' ( ) & * + , - Vol.35No.2
2024 $ 4 % JournalofWaterResources&WaterEngineering Apr.,2024
DOI:10.11705/j.issn.1672-643X.2024.02.08
'ÅaD<y8 VMD-LSTM2
34»{\3|ST
1 1 1,2 1,2 1,2 3
3tu ,SOOMROShanEHyder, Y v , YwI , > x , Ryz
(1. [\:+ 2]^_`+a, EF bc 443002;2. [\:+ 2345¾4^»2EFh9xyz{,
EF bc 443002;3. ´]«E´»2{, EF ´] 433000)
B C: °ªÍ, ¼ÆùþüÄÅPЫ, ãT265¦³¸$u, Ç®ãTmbÈÚã·¦P
c#$。¹O± LSTM( b¿Èµ¶`ñab) ¿ÀËÚãµ*¦·¸8Ç)´¿, 7J_ãT¹./ÝT,
ßÿùþ、 Æ«8ã¡¥n, ]@ VMD( ¿G) üÚ;|z, ûo;¡Ë LSTM¿ÀD+6
¦Ä?, 7¹s¿À7=H¦1¦; W, ,ƾ VMD^ LSTM¿À¦³j)ÙË¿Àcü¦
?@, ÿHG"ÈüY¦ VMD-LSTMÿÚã)¿Ù。ëìíî:VMD-LSTM)¿Àkö÷O±¿Ç
c, Ó¿À¨)r¹¾; ÊÓÔÚ;|z¦ VMD-LSTM)¿À³¬´(¶[O&¿Àc, ×
«¿À¦¨Ë)rº¶。ÓÔÚ;|z¦ VMD-LSTM)¿À¦³j)ÙË"m, Ë7
=、 7G©¶· VMDG«Ë7=¡¶·H¦ D )~¦¿Çñìÿó, ý 60 #¿Ç£Ö¦ NSE©¹
1
0.98 7»«×; ®, ~ D ¦kGܵ«0!"ÚãËÔ LSTM¿À¦?@N"ùþüÆ«
1
:。Ñ./ë}k¹ãT2QR»2OPSkܦmbÈÚã¿Çjì。
DEF: Ú;|z; ¿G; b¿Èµ¶`ñab; Úã¿Ç; J_ãT
GHIJK:P338 LMNOP:A LQRK:1672643X(2024)02007112
SimulationofmonthlyrunoffbyVMD-LSTM coupled
modelbasedoncorrelationtesting
1,2
1,2
1
1
LIUShenghong,SOOMROShanEHyder,LIYing ,LIYinghai ,
3
1,2
CHENGXiong ,YANGShaokang
(1.CollegeofHydraulic&EnvironmentalEngineering,ChinaThreeGorgesUniversity,Yichang443002,China;
2.HubeiKeyLaboratoryofHydropowerEngineeringConstructionandManagement,Yichang443002,China;
3.XiantaoRiverEmbankmentAdministration,Xiantao433000,China)
Abstract:Inrecentyears,thefrequentoccurrenceofextremeheavyrainfallanddroughteventshasin
tensifiedtheuncertaintychangesofhydrologicalprocessesinthebasin,makingitmoredifficultforthe
predictionofmediumandlongtermrunoffinthebasin.InordertoimprovetheabilityofLSTM (long
shorttermmemory)modelincapturingandfittingtemporalchangesinrunoff,takingtheBoyangRiver
Basinasthestudyarea ,wecollectedthedataofmonthlyrainfall,evaporationandflowforthesimula
tion.Then ,VMD(variationalmodaldecomposition)andcorrelationtestingareusedtoeliminatethein
terferenceofirrelevantfrequencycomponentsontheregularlearningofLSTM modelforthepurposeof
modelinputoptimization.Inaddition,theeffectsofdifferentcouplingmethodsbetweenVMDandLSTM
modelontheaccuracyandstabilityofthemodelareconsidered ,bywhichtheVMD-LSTMmonthlyrun
offcoupledmodelwhichtakestheadvantagesofbothisfinallyselected.Thesimulationresultsshowthat
theVMD-LSTM coupledmodelcansignificantlyimprovethesimulationaccuracybutitislackingin
!"#$:20230516; %$:20231121
'()*: qw,-*+ÓÛ01(52179018,51909010);qw9x.«£æ~ö(2022YFC32039023);]Zbâ^2
3*+EFh9xyz{( mqbâ3¿±Ð)dKd) äÇÓÛ01(ZH2002000103);bâ*+aäÇ.
/ÓÛ01(CKWV2021889/KY)
+,-.: ˲°(1999—), D, â³~I, sK./L, MN./ö¹26¿À82678)。