Page 49 - 《水资源与水工程学报》2023年第4期
P. 49
!34 " ! 4 # & ' ( ) & * + , - Vol.34No.4
2023 $ 8% JournalofWaterResources&WaterEngineering Aug.,2023
DOI:10.11705/j.issn.1672-643X.2023.04.06
'az{~:;<rx½=2d
GRU/LSTM6¾#/>?>æ
1
2
3
VZ[ , N \ , ]^_ , V ` 4
(1. ®ð@A@CDÜ, ®ð 3 650106;2. VÓ4@CD%H?øÿ[5, ®ð V 677000;3. ®ðAÏÑ
@øÜ, ®ð AÏ 663000;4. EÕÞÀ;<ËÌP,6ÿ[53·5, ®ð 3 650041)
H I: ?uX@¡âfg[ÁZ, uh¼½'x(¨U?(WPT) WU¦qV8p(CSA)、 78qV(CO)
8pWÏ9:qV(MGO) 8p^qV/[áÔÃa3(GRU)、 8ñNÉÊænµ(LSTM) ^[qr。öJ
WPT o pHÆ、DO、COD 、NH—Nâfg}¶V, ¥åbÏÇÈlûY^Ûfg·Z; /ì.
Mn
3
;o CSA、CO、MGO8p®, J CSA、CO、MGO·kq GRU、LSTMõûù, Ls WPT-CSA-GRU、WPT-CO-
GRU、WPT-MGO-GRU、WPT-CSA-LSTM、WPT-CO-LSTM、WPT-MGO-LSTMqr; ±ôJéLs^qro
pHÆB DO、COD 、NH—NeZv·Z}¶[W&¶, ÄLs WPT-GRU、WPT-LSTMW WPT-CSA-SVM、
3
Mn
WPT-CO-SVM、WPT-MGO-SVM qrýoº·1qr, ήð3ªÏëÎ?«, î pHÆB DO、
COD 、NH—NeZ[oqr}¶oÛ!。¯ï:WPT-CSA-GRU、WPT-CO-GRU、WPT-MGO-GRU、
3
Mn
WPT-CSA-LSTM、WPT-CO-LSTM、WPT-MGO-LSTMqro¾« pHÆB DO、COD 、NH—NeZ^[ÁZ
3
Mn
qx/poºqr, þÿû¥^[., /¢öÎ WPT-CSA-GRU、WPT-CO-GRU、WPT-MGO-GRUqr^
[ÁZ±X;CSA、CO、MGOQÿ.Óq GRU、LSTMõûù, _`uX GRU、LSTM[lQ; é¶L^ 6 ½qr[
ÁZXÌ8Çq(, ´ÿ.^@¡âfg[qr, µ?{|@¡[;<uÕûü。
JKL: @¡[; /[áÔÃa3; 8ñNÉÊænµ; (¨U?; U¦qV8p; 78qV8p;Ï9
:qV8p
MNOPQ:TV213.4 7RSTU:A 7VWQ:1672643X(2023)04004509
PredictionofGRU/LSTM waterqualitytimeseriesbasedondataprocessing
andoptimizationofseveralswarm intelligencealgorithms
3
2
4
1
YANGPinghong,HUAo,CUIDongwen,YANGJie
(1.HydrologyandWaterResourcesBureauofYunnanProvince,Kunming650106,China;2.LincangRuntingWaterResources
TechnologyServiceCo. ,Ltd.,Lincang677000,China;3.WenshanPrefectureWaterAffairsBureauofYunnanProvince,Wenshan
663000,China;4.KunmingBranchofCRSCResearchandDesignInstituteGroupCo.,Ltd.,Kunming650041,China)
Abstract:Toimprovethepredictionaccuracyofwaterqualitytimeseries,atypeofpredictionmodelbased
onwaveletpackettransform (WPT),chameleonoptimizationalgorithm (CSA),cheetahoptimization
(CO)algorithmandmountaingazelleoptimization(MGO)algorithmisproposedtooptimizethegatedre
currentunit (GRU)andlongshorttermmemorynetworks(LSTM).Firstly,thetimeseriesofpH,DO,
COD andNH—N arestabilizedbyWPT,bywhichseveralregularsubsequencecomponentsareob
Mn
3
tained.Secondly ,theCSA,COandMGOalgorithmsarebrieflyintroducedandthenappliedtotheoptimi
zationofthehyperrparametersofGRUandLSTM ,bywhichtheWPT-CSA-GRU,WPT-CO-GRU,
WPT-MGO-GRU,WPT-CSA-LSTM,WPT-CO-LSTM,WPT-MGO-LSTM modelsareestab
!"#$:20221127; %$:20230227
'()*: !"&Ñ;'ÌÊ)*(2021YFC3200903); !"#$%&'()*(51809288); ¢!@J@Î%&;<P'
u%;øX)*(WR0145B022021)
+,-.: Qôq(1967—), ., ®ðeô3, XýNO:, @É@2#B@CDFÕNý。
¥¦+,: 2ÐA(1978—), ., ®ð¾3, &5, AXýNO:, @É@CDÍFÕB´Q8p@A@CDW
#¢^¬;<Ný。