m''!'.metnet.imd.gov.in/mausamdocs/536410.pdf · 2019. 3. 18. · nre presented graphically and in...
6
M ausam, (1985). 36. 4, 463-4ols 555'1 6 : 556 .5, Flow discharge prediction of river Tigris using self tun ing predictor s* A. K. SINHAt indian Institute of Technology, Ne .... Delht and SA DIQ JA SS IM University of Basrah, Basrah, Iraq (Received 2 April 1981) -l'I'li</ ..... ii """. ilfn.-< oR 1"m''''' ror: mrr.r..-.rr tm; ii if'Im qft om ';: lI"<Tl; funl" qft llT'Tfuf ii m "'l""l<r' '1 lrt<.!" fl;trr 'l'lT t,l'Iii """'" <w.fr<it "" F,;or 'l'lT t I 'IT':.-.:n: ". ""lrnt 'I'T':'f.i/€mro (,) "l""" 'fl'r l1frn-<. (2) irfTro:. ( 3) llWlIl'O lIFt"'" 'l'T '3'l1ll<r rn.-mT 1I'lT" .. "",,) .-mT 11ft ",. Oil fir<;r" fl;trr = t I ..._ lf tq1 <iT ']'1T.l<r ""i; fO'lfui om "' 1I'I'IlI 'f.n;, 'ITf'ff..-.rr W .. lIT'lf.H '"!""'."" "" f>R'lll t I llPl,ffil.a ""'" mor ii it f.r<>il 'li't '1l1T t, T'f"T "' f>R filf1l'"""lrnt 'lRvrr'I m '" opt ii..-.rr r "" ii "'¥ 'lll I AUSTRACf. The pape r pre sent s development of recursive sel f-adaptive prediction algo rithms called the Self-tuning predictors using some common estimation techniques and thei r applica tion to predicti on of flow discharge of river Tigris at Baghdad, Iraq. Four kinds of predict ors. name lv , the Icast square predictor, the minimum variance predictor, predictor using stochastic approx ima tio n and a two stage predictor have been developed. Using available data for the river Tigris. predict ion results have bee n o btai ned Io r ave rage daily discharge. ave rage monthl y discharge ami ave rage yearly dis- charge. In each typt ( I t" predicti on , a number of mode ls have been tried. The var io us predicti on results nre pre sen ted graphicall y and in tabular forms for comparison. 1. Introduction The application of modern recursive estimation tech- niques, such as least squares and Kalman filtering ( Gra u- pe 1976) is well known in the areas of comm unication and cont rol. In co mmunication. the techn iques are used generally to filter out signals fro m noisy info rma tions. In cont rol, these arc used normally for state and para- meter estimatio n of various processes to be controlled. In the recent years re searc he rs have also used these techniques to solve the problem of modelling and pre- diction in o ther areas, such as educational system (Sinha and Sin gh 1973). weather process (Sinha and Sharma 1975) and river flows ( Rao and Kash yap 1974). In the present paper. the authors have tried 10 usc a unified approach to the pro blem of recursive predic- tion in such processes by developing four self ada ptive prediction algorithms called the self tuning predictors (Willenmar k 1974) based on some common estimation techniques like the least square, minimum variance. stochastic approximation (Graupe 1976) and two stage estima tion (Prasad, Sinha and Mahalanabis 1977) algorithm. These have been used to process real flow discha rge dat a of rive r Tigris at Baghdad (Iraq) to obtain three kinds of pred iction s. namely. the average daily dis- charge prediction one day in adva nce the average mono thly discharge prediction one month in advance an d the average yea rly discharge prediction one year in advance. Th e models devel oped are stochastic difference equat ion s with or without deterministic input term s and a lso with sinosoidally varying terms in some cases to take into accou nt the periodic nature of the river flow data. The approximate model orders are obtained by o rdinary auto-correlation and cross co rrelation tests (S inh a and Sharm a (975) and the models are valida ted by obtaining the error of prediction. 2• •"redictor algorithms In this sectio n we consider the prob lem of prediction of a process variable. such as river ft ow discharge and develop various kinds of prediction algorithms . "The paper wa s present ed in the Sym nosiurn "I ndo-French School on recent advance s in Computer techn ique s in Meteorology, Biomechanics and Appl ied Systems" held at J.I.T.. New De lhi. February 1980. t p emlsed in A.I. Kunishk u air-crush on 23 June 1985. (463,
Transcript of m''!'.metnet.imd.gov.in/mausamdocs/536410.pdf · 2019. 3. 18. · nre presented graphically and in...
M ausam, (1985). 36. 4, 463-4ols
555 '1 6 : 556 .5,
Flow discharge predict ion of river Tigris usingself tun ing predictors*
A. K. SINHAt
indian Institute of Technology, Ne .... Delht
a nd
SA DIQ JA SSIM
University of Basrah, Basrah, Iraq
(Rece ived 2 April 1981)
~ -l'I'li</..... ii """.~ ilfn.-<~ oR1"m''''' ror: lf~ lfl~mr ~"iiF<"l "mrr.r..-.rr tm;iiif'Imqft~ om ';: lI"<Tl; funl" qft llT'Tfuf iim "'l""l<r''1 lrt<.!" fl;trr 'l'lT t,l'Iii~g """'"~ <w.fr<it "" ~liprF,;or'l'lT t I 'IT':.-.:n: ".""lrnt 'I'T':'f.i/€mro~"'. ( ,) "l""" 'fl'r l1frn-<. (2) "i"'f'l lfT~ irfTro:. ( 3) llWlIl'O lIFt"'"'l'T '3'l1ll<r rn.-mT~ 1I'lT" .. ~r "",,) .-mT 11ft",. Oil fir<;r" fl;trr = t I ..._ lftq1 <iT ']'1T.l<r ""i; fO'lfui om"' 1I'I'IlI 'f.n;, 'ITf'ff..-.rr W ..~", ,, lIT'lf.H '"!""'."""" f>R'lll t I llPl,ffil.a""'" mor ii ~~ it f.r<>il 'li't~'1l1T t , T'f"T "' f>R filf1l'"""lrnt'lRvrr'Im'" opt ii..-.rr ""~r "" ii"'¥~ 'lll ~ I
AUSTRACf. The paper present s development of recursive sel f-adaptive predict ion algo rithmscalled the Self-tuning predicto rs using some common estimatio n techniques and thei r application topredicti on of flow discharge of river Tigris at Baghdad, Iraq. Four kinds o f predictors. name lv , theIcast square predictor, the minimum variance predictor, predictor using stochast ic approx imation anda two stage predicto r have been deve loped. Using avai lable data fo r the river Tigris. predict ion resultshave been o btai ned Io r ave rage daily discharge. average monthl y discharge ami ave rage yearly discharge. In each typt ( I t" prediction, a number of mode ls have been tried. The var ious predicti on resultsnre pre sen ted graphicall y a nd in tabular forms for comparison.
1. Introduction
The application of modern recursive estimation techniques, such as least sq uares and Kalma n filter ing (Graupe 1976) is well kno wn in the a reas of communicationand control. In communica tion . the techn iques a re usedgenerally to filter o ut signa ls fro m no isy informations.In cont rol , these a rc used no rma lly for sta te and par ameter estimatio n of vario us processes to be controlled.In the recent years researchers have also used thesetech niqu es to solve the problem of modelling and prediction in other areas, such as educational system(Sinha and Singh 1973). weather process (Sin ha andSharma 1975) and river flows (Rao and Kashyap 1974).
In the present paper. the autho rs have tried 10 usca unified a pproac h to the problem of recursive prediction in such processes by deve loping four self ada pt iveprediction a lgo ri thms ca lled the self tu ning predictors(Willenmar k 1974) based on so me common estimat iontechniques like the least square, minimum variance.
stoc has tic approxima tio n (G ra upe 1976) and two stageestimation (Prasad, Sinha and Mahalanabis 1977)algo rithm. These have been used to process real flowdischa rge dat a of rive r Tigris at Baghdad (Iraq) to obtainthree kinds of pred iction s. namely. the average daily d ischa rge pred iction one day in advance the average mo nothl y d ischarge prediction one month in advance an d theaverage yea rly discharge prediction one year in advance.The models developed are stochas tic difference equation swith or witho ut de terministic input terms and a lsowith sinosoidally varying terms in some cases to takeinto accou nt the periodic nature of the river flow data.The a pproximate model orders are obtai ned by ordinarya uto-correla tio n a nd cro ss co rrelat ion tests (Sinh aand Sharma (975) and the models ar e valida ted byobtai ning the error of prediction.
2• •"redictor algorithms
In th is sectio n we con sider the problem of predictionof a process variable. such as river ftow discharge anddevelop va rio us kinds of prediction algorithms.
"The paper wa s present ed in the Sym nosiurn "Indo- French School on recent advances in Computer techniques in Meteorology,Biomechanics and Appl ied Systems" held at J.I.T.. New De lhi. February 1980.
t p emlsed in A.I. Kunishku ai r-crush on 23 June 1985.(463,
