Page 184 - 《水资源与水工程学报》2023年第2期
P. 184
!34 " ! 2 # & ' ( ) & * + , - Vol.34No.2
2023 $ 4 % JournalofWaterResources&WaterEngineering Apr.,2023
DOI:10.11705/j.issn.1672-643X.2023.02.23
'Ö EEMDLSTMARIMAhY3ëd>cå÷AB
ö÷ø, 0ù, jú
( Á\& 010É&`, äÐ LÊ 210098)
C D: Ôu~]F/êZLM4Ëá¢.。ÛÔ ílý7ëì34, u5Ôè±
£wlúúX'¸、 ¶hÄ,¹5, Gf'“ ³" 4Ý ¢” RzÝtÀ]FÔtu EEMD
LSTMARIMA®。mò÷fÌÍxo³"(EEMD) $±£wl¹º®, Ük®ä*þã-ÍEI
(LSTM) $®¹º³tu, ڱǵ³#sTF¬E7(ARIMA) ¥µ/{, ¢ LSTMÓ
ARIMAtuÇÈ, 4ÝEtu®。ÐeuvO]F23Gû, Ý-:F±Õ 2
Â^®u5wuÔèwlG6C$Îfx¼。ÇÈÉÊ: (mê. LSTM®Ó ARIMA®, E
®¬$[µ MAE、 ¬E[µ MSE、 ¬EÜ[µ RMSE¬G 3 ½®iÔ9è, tuuÊà`·ý
2 ½®, ¿®G]FÔu;tu³¿®¯À*{ñ。
>EF: ]F; Ôtu; ÌÍxo³"; ®ä*þã-ÍEI; µ³#sTF¬
GHIJK:TV641 LMNOP:A LQRK:1672643X(2023)02018006
Seepageporewaterpressurepredictionmodelofearth-rock
damsbasedonEEMDLSTMARIMA
CENWeijun,WANGXiaoxin,JIANGMinghuan
(CollegeofWaterConservancyandHydropowerEngineering,HohaiUniversity,Nanjing210098,China)
Abstract:Seepagemonitoringisoneoftheimportantcontentsofseepagesafetyevaluationofearth-rock
dams.Theseepageporewaterpressureisaffectedbymultipleexternalfactors ,sothetimeseriesofseep
ageporewaterpressureatmeasuringpointsisoftencharacterizedbynonstationarityandlocalabrupt
changes.Regardingtothis ,theEEMDLSTMARIMAmodelforseepageporewaterpressureprediction
ofearth-rockdamsisconstructedaccordingtotheconceptofdecomposition-reconstruction-combina
tion.Firstly ,thetimeseriesfeaturesaredecomposedbytheensembleempiricalmodedecomposition
(EEMD),andtheextractedfeaturecomponentsarepredictedbythelongshorttermmemory(LSTM)
neuralnetwork.Atthesametime,theresidualerroriscorrectedbytheautoregressiveintegratedmoving
average (ARIMA),andtheimprovedpredictionmodelisreconstructedbycombiningthepredictionre
sultsofLSTMandARIMA.Takinganearth-rockdamonadeepoverburdenasanexample,themeas
uredseepageporewaterpressureseriesoftwotypicalmeasuringpointsbehindthecutoffwallofthemain
riverbeddamareselectedastheresearchobjectsforapplicationverification.Theresultsshowthat ,com
paredwiththesingleLSTM modelandARIMAmodel ,themeanabsoluteerror,themeansquareerror
andtherootmeansquareerroroftheproposedpredictionmodelarethesmallest,andthepredictionac
curacyoftheproposedmodelisobviouslysuperiortotheothertwomodels.Therefore ,theproposedmod
elcanprovideanewapproachforaccuratepredictionandanalysisofseepageporewaterpressureof
earth-rockdams.
Keywords:earth-rockdam;porewaterpressureprediction;ensembleempiricalmodedecomposition(EE
MD);longshorttermmemory(LSTM)neuralnetwork;autoregressiveintegratedmovingaverage(ARIMA)
!"#$:20220722; %$:20220914
'()*: uv#$%&'()*(51979089、51679073)
+,-.: á$(1977—), ë, ^äâL@, AB, PQ, ABDST, 6CEFG]Fù_23/êZ。