{"id":3504,"date":"2017-06-12T08:36:24","date_gmt":"2017-06-12T08:36:24","guid":{"rendered":"http:\/\/www.vtei.cz\/?p=3504"},"modified":"2024-07-16T12:10:06","modified_gmt":"2024-07-16T11:10:06","slug":"stochastic-management-of-the-open-large-water-reservoir-with-storage-function-with-using-evolution-algorithm","status":"publish","type":"post","link":"https:\/\/www.vtei.cz\/en\/2017\/06\/stochastic-management-of-the-open-large-water-reservoir-with-storage-function-with-using-evolution-algorithm\/","title":{"rendered":"Stochastic management of the open large water reservoir with storage function with using evolution algorithm"},"content":{"rendered":"

<\/i> This article is available in Czech only. For translation or more information on this topic, please contact author.<\/h4>\n

 <\/p>\n

Souhrn<\/h2>\n

V\u00a0\u010dl\u00e1nku popsan\u00fd model pou\u017e\u00edv\u00e1 stochastickou p\u0159edpov\u011b\u010f r\u016fzn\u00e9 d\u00e9lky. P\u0159edpov\u011b\u010f je v\u017edy krat\u0161\u00ed ne\u017e 1\u00a0rok. Pro adaptivn\u00ed \u0159\u00edzen\u00ed je pou\u017eita pouze prvn\u00ed hodnota \u0159ady \u0159\u00edzen\u00fdch odtok\u016f. \u0158ada p\u0159edpov\u011bzen\u00fdch p\u0159\u00edtok\u016f vody do n\u00e1dr\u017ee je transformov\u00e1na na \u0159adu \u0159\u00edzen\u00fdch odtok\u016f vody z\u00a0n\u00e1dr\u017ee. Stochastick\u00e9 \u0159\u00edzen\u00ed pracuje s\u00a0rozptylem \u0159\u00edzen\u00fdch odtok\u016f. Stochastick\u00e9 \u0159\u00edzen\u00ed pracuje s\u00a0v\u011bj\u00ed\u0159em mo\u017en\u00fdch hodnot. \u010cl\u00e1nek popisuje konstrukci a\u00a0vyhodnocen\u00ed adaptivn\u00edho stochastick\u00e9ho modelu, kter\u00fd pou\u017e\u00edv\u00e1 pro optimalizaci evolu\u010dn\u00ed algoritmy. Model byl pou\u017eit pro stochastick\u00e9 \u0159\u00edzen\u00ed z\u00e1sobn\u00ed funkce n\u00e1dr\u017ee. P\u0159i zad\u00e1n\u00ed vstup\u016f (p\u0159edpov\u011bzen\u00e9 hodnoty p\u0159\u00edtok\u016f) je modelem pro zvolenou pravd\u011bpodobnost p\u0159ekro\u010den\u00ed vypo\u010dten \u0159\u00edzen\u00fd odtok vody z\u00a0n\u00e1dr\u017ee. Model byl testov\u00e1n i\u00a0validov\u00e1n na um\u011bl\u00e9 vodn\u00ed n\u00e1dr\u017ei (profil B\u00edlovice nad Svitavou) z\u00a0d\u016fvodu dostupnosti dat. Model \u0159\u00eddil n\u00e1dr\u017e logicky a\u00a0p\u0159i zv\u00fd\u0161en\u00ed celkov\u00e9ho po\u010dtu p\u0159edpov\u011bd\u00ed (z\u00a0300 na 500) se zlep\u0161il celkov\u00fd pr\u016fb\u011bh \u0159\u00edzen\u00ed. Dal\u0161\u00ed zvy\u0161ov\u00e1n\u00ed celkov\u00e9ho po\u010dtu p\u0159edpov\u011bd\u00ed nep\u0159ineslo v\u00fdrazn\u00e9 zlep\u0161en\u00ed. Model vyu\u017e\u00edvaj\u00edc\u00ed evolu\u010dn\u00ed algoritmy by pot\u0159eboval velk\u00e9 mno\u017estv\u00ed strojov\u00e9ho \u010dasu, a\u00a0proto byly v\u00fdpo\u010dty prov\u00e1d\u011bny v\u00a0clusteru.<\/p>\n\"P8280190_edit\"<\/a>\n

\u00davod<\/h2>\n

V\u00a0sou\u010dasn\u00e9 dob\u011b jsou v\u00a0praxi pou\u017e\u00edv\u00e1ny p\u0159i \u0159\u00edzen\u00ed z\u00e1sobn\u00ed funkce n\u00e1dr\u017ee p\u0159ev\u00e1\u017en\u011b deterministick\u00e9 modely. V\u00fdhodou pou\u017eit\u00ed deterministick\u00fd model\u016f je pouze jeden v\u00fdstup (hodnota) \u0159\u00eddic\u00ed veli\u010diny, av\u0161ak p\u0159i pou\u017e\u00edv\u00e1n\u00ed pouze jedn\u00e9 hodnoty doch\u00e1z\u00ed d\u00edky siln\u00e9mu zjednodu\u0161en\u00ed problematiky ke ztr\u00e1t\u011b p\u0159esnosti \u0159\u00edzen\u00ed \u010di myln\u00e9mu vyhodnocen\u00ed situace (p\u0159edpov\u011b\u010f se m\u016f\u017ee v\u00fdrazn\u011bji odli\u0161ovat od skute\u010dnosti, kter\u00e1 byla p\u0159edpokl\u00e1d\u00e1na) [1]. Oproti tomu stochastick\u00e9 \u0159\u00edzen\u00ed pracuje s\u00a0ur\u010dit\u00fdm rozptylem hodnot \u0159\u00eddic\u00edch pr\u016ftok\u016f s\u00a0dan\u00fdm pravd\u011bpodobnostn\u00edm rozd\u011blen\u00edm, a\u00a0proto doch\u00e1z\u00ed k\u00a0v\u00fdrazn\u011b lep\u0161\u00ed aproximaci skute\u010dn\u00e9 problematiky \u0159\u00edzen\u00ed. V\u00fdhodou stochastick\u00e9ho \u0159\u00edzen\u00ed je v\u00fdb\u011br mo\u017enosti \u0159\u00edzen\u00ed pro danou pravd\u011bpodobnost sc\u00e9n\u00e1\u0159e. V\u00fdb\u011br pravd\u011bpodobnost\u00ed n\u00e1m poskytne v\u011bj\u00ed\u0159 mo\u017enost\u00ed.<\/p>\n

Pro stochastick\u00e9 \u0159\u00edzen\u00ed m\u016f\u017eeme p\u0159i adaptivn\u00edm zp\u016fsobu \u0159\u00edzen\u00ed vyu\u017e\u00edt metodu Monte Carlo [2]. Jej\u00ed vyu\u017eit\u00ed v\u00a0oblasti vodohospod\u00e1\u0159sk\u00e9ho \u0159e\u0161en\u00ed n\u00e1dr\u017e\u00ed nar\u00e1\u017e\u00ed na omezen\u00ed z\u00a0hlediska v\u00fdpo\u010detn\u00ed techniky (velk\u00e9 n\u00e1roky na strojov\u00fd \u010das). Pro vyhodnocen\u00ed opakovan\u00fdch n\u00e1hodn\u00fdch stav\u016f je zapot\u0159eb\u00ed minim\u00e1ln\u011b 300 opakov\u00e1n\u00ed v\u00fdpo\u010dtu. Adaptivn\u00ed \u0159\u00edzen\u00ed s\u00a0pou\u017eit\u00edm metody Monte Carlo a\u00a0vyu\u017eit\u00edm tradi\u010dn\u00edch algoritm\u016f pro optimalizaci (krit\u00e9riem optimalizace budeme v\u00a0textu ch\u00e1pat odchylku mezi nadlep\u0161en\u00fdm a\u00a0\u0159\u00edzen\u00fdm odtokem, kter\u00e1 bude minimalizov\u00e1na) je vhodnou volbou pro stochastick\u00e9 \u0159\u00edzen\u00ed. Av\u0161ak v\u00fdpo\u010detn\u00ed doba je pro b\u011b\u017en\u00e9 po\u010d\u00edta\u010de ne\u00fanosn\u00e1. Velk\u00e9 po\u017eadavky na v\u00fdpo\u010detn\u00ed techniku jsou zp\u016fsobeny p\u0159edev\u0161\u00edm optimalizac\u00ed, kter\u00e1 je pou\u017eita v\u00a0ka\u017ed\u00e9m \u010dasov\u00e9m kroku, ve kter\u00e9m opakovan\u011b korigujeme \u0159\u00edzen\u00ed (vych\u00e1z\u00ed z\u00a0principu adaptivity). Optimalizace vyu\u017e\u00edv\u00e1 tradi\u010dn\u00ed algoritmy (Diferenci\u00e1ln\u00ed evoluci). V\u00fdpo\u010dty byly prov\u00e1d\u011bny v\u00a0clusteru, proto\u017ee n\u00e1roky na strojov\u00fd \u010das u\u00a0b\u011b\u017en\u00e9ho po\u010d\u00edta\u010de by byly ne\u00fanosn\u00e9. C\u00edlem je navrhnout algoritmus, kter\u00fd bude \u00fa\u010dinn\u011b minimalizovat ekonomick\u00e9 ztr\u00e1ty vznikl\u00e9 z\u00a0nedod\u00e1vky vody (nespln\u011bn\u00e9 hodnoty odtoku vody). M\u011bl\u010d\u00ed av\u0161ak del\u0161\u00ed poruchy (nap\u0159. m\u00edrn\u00e9 omezen\u00ed v\u00fdroby) jsou ekonomicky v\u00fdrazn\u011b p\u0159\u00edzniv\u011bj\u0161\u00ed ne\u017e hlub\u0161\u00ed a\u00a0krat\u0161\u00ed poruchy (nap\u0159. v\u00fdrazn\u00e9 omezen\u00ed, zastaven\u00ed v\u00fdroby, nedostate\u010dn\u00fd pr\u016ftok vody v\u00a0toku pro sm\u011bsn\u00fd pom\u011br s\u00a0vodou vypou\u0161t\u011bnou z\u00a0\u010dist\u00edrny odpadn\u00edch vod atd.) Pro sestaven\u00ed a\u00a0pr\u00e1ci s\u00a0modelem byl pou\u017eit program Matlab 2013.<\/p>\n

Model<\/h2>\n

Model pou\u017e\u00edv\u00e1 princip adaptivity a\u00a0jeho j\u00e1drem je optimaliza\u010dn\u00ed modul vyu\u017e\u00edvaj\u00edc\u00ed diferenci\u00e1ln\u00ed evoluci (DE) [3], kter\u00e1 pou\u017e\u00edv\u00e1 t\u0159i rodi\u010de. Modelu je poskytnuta \u0159ada budouc\u00edch p\u0159\u00edtok\u016f vody do n\u00e1dr\u017ee (p\u0159edpov\u011bzen\u00e9 pr\u016ftoky), na z\u00e1klad\u011b kter\u00fdch by m\u011bl b\u00fdt model schopen naj\u00edt nejlep\u0161\u00ed mo\u017en\u00e9 \u0159e\u0161en\u00ed podle kriteri\u00e1ln\u00ed funkce, pokud by p\u0159\u00edtoky vody do n\u00e1dr\u017ee nebyly zat\u00ed\u017eeny nejistotami. Nejistota spojen\u00e1 s\u00a0p\u0159edpov\u011bd\u00ed m\u016f\u017ee v\u00e9st k\u00a0jin\u00e9mu pr\u016fb\u011bhu \u0159\u00edzen\u00ed, ne\u017e kter\u00e9 by poskytla 100% p\u0159edpov\u011b\u010f (re\u00e1ln\u00e1 \u0159ada) [4].<\/p>\n\"Kozel-1\"<\/a>\n

Obr. 1. \u0158\u00eddic\u00ed algoritmus modelu
\nFig. 1. Control algorithm of model<\/h6>\n

Uveden\u00e1 nejistota m\u016f\u017ee zp\u016fsobit rychlej\u0161\u00ed pr\u016fb\u011bh pr\u00e1zdn\u011bn\u00ed z\u00e1sobn\u00edho objemu vody n\u00e1dr\u017ee, ne\u017e jak by bylo pot\u0159ebn\u00e9 (n\u00e1dr\u017e je pr\u00e1zdn\u011bna po skoc\u00edch, ne plynule). Rychlost (gradient) pr\u00e1zdn\u011bn\u00ed n\u00e1dr\u017ee je hl\u00edd\u00e1na modelem, a\u00a0pokud rychlost pr\u00e1zdn\u011bn\u00ed nab\u00fdv\u00e1 v\u00fdrazn\u00fdch hodnot ve dvou \u010dasov\u00fdch \u00fasec\u00edch po sob\u011b jdouc\u00edch, model odtok vody z\u00a0n\u00e1dr\u017ee sn\u00ed\u017e\u00ed o\u00a0pom\u011br mezi rychlost\u00ed pr\u00e1zdn\u011bn\u00ed a\u00a0maxim\u00e1ln\u00ed hodnotou z\u00e1sobn\u00edho objemu pro n\u00e1sleduj\u00edc\u00ed \u010dasov\u00fd krok. Druh\u00fdm d\u016fvodem zaveden\u00ed pomocn\u00e9ho algoritmu je, \u017ee metoda DE pracuje s\u00a0cel\u00fdm z\u00e1sobn\u00edm objemem n\u00e1dr\u017ee (v\u010detn\u011b p\u0159edpov\u011bdi). Pokud bude p\u0159edpov\u011b\u010f vy\u0161\u0161\u00ed ne\u017e re\u00e1ln\u00fd pr\u016ftok, kter\u00fd nastane, m\u016f\u017ee se p\u0159i dlouho trvaj\u00edc\u00edm suchu vy\u010derpat velmi rychle po skoc\u00edch cel\u00fd z\u00e1sobn\u00ed prostor n\u00e1dr\u017ee a\u00a0n\u00e1sledn\u011b dojde ke vzniku velmi hlubok\u00e9 poruchy. Poruchy p\u0159i \u0159\u00edzen\u00ed dod\u00e1vky vody by m\u011bly b\u00fdt ide\u00e1ln\u011b m\u011bl\u010d\u00ed a\u00a0del\u0161\u00ed ne\u017e krat\u0161\u00ed a\u00a0hlub\u0161\u00ed. Model provede \u0159\u00edzen\u00ed pro v\u0161echny p\u0159edpov\u011bdi, sestav\u00ed empirickou \u010d\u00e1ru p\u0159ekro\u010den\u00ed \u0159\u00edzen\u00fdch odtok\u016f a\u00a0ode\u010dte odtok vody z\u00a0n\u00e1dr\u017ee pro po\u017eadovanou pravd\u011bpodobnost. Pot\u00e9 je hodnota odtoku korigov\u00e1na a\u00a0podle rovnice (1) je spo\u010dtena nov\u00e1 hodnota z\u00e1sobn\u00edho objemu. Kriteri\u00e1ln\u00ed funkce je suma druh\u00fdch mocnin odchylek \u0159\u00edzen\u00e9ho odtoku vody z\u00a0n\u00e1dr\u017ee od \u0159\u00edd\u00edc\u00edho odtoku vody z\u00a0n\u00e1dr\u017ee. Na obr. 1<\/em> je uveden hlavn\u00ed algoritmus \u0159\u00edzen\u00ed n\u00e1dr\u017ee.<\/p>\n\"Kozel-vzorec1\"<\/a>\n\n\n\n\n\n\n
Kde\u00a0<\/em><\/td>\nQ<\/em>i <\/sub>[m<\/em>3<\/sup>\/s]\u00a0<\/em><\/td>\n\u00a0<\/em>je<\/td>\nvypo\u010dten\u00fd odtok vody pro zvolenou pravd\u011bpodobnost,<\/td>\n<\/tr>\n
<\/td>\nQp<\/em>i <\/sub>[m<\/em>3<\/sup>\/s]\u00a0\u00a0<\/em><\/td>\n<\/td>\n\u00a0\u00a0<\/em>re\u00e1ln\u00fd p\u0159\u00edtok vody do n\u00e1dr\u017ee,<\/td>\n<\/tr>\n
<\/td>\nV<\/em>i <\/sub>[m<\/em>3<\/sup>\/s]\u00a0\u00a0<\/em><\/td>\n<\/td>\nobjem vody v\u00a0n\u00e1dr\u017ei na po\u010d\u00e1tku \u0159e\u0161en\u00e9ho kroku,<\/td>\n<\/tr>\n
<\/td>\nV<\/em>i+1 <\/sub>[m<\/em>3<\/sup>\/s]<\/em><\/td>\n<\/td>\nna konci \u0159e\u0161en\u00e9ho kroku.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

Korekce \u0159\u00edzen\u00ed<\/h2>\n

P\u0159i pou\u017eit\u00ed metody DE jsou \u0159\u00edzen\u00e9 odtoky vody \u010dasto velmi rozkol\u00edsan\u00e9, a\u00a0proto bylo p\u0159istoupeno ke korekci \u0159\u00edzen\u00e9ho odtoku vody. Korekce vych\u00e1z\u00ed z\u00a0my\u0161lenky\u00a0metody konjugovan\u00fdch gradient\u016f. \u0158\u00edzen\u00fd odtok vody z\u00a0n\u00e1dr\u017ee je zpr\u016fm\u011brov\u00e1n s\u00a0p\u0159edchoz\u00edmi kroky \u0159\u00edzen\u00ed pro danou pravd\u011bpodobnost \u0159\u00edzen\u00e9ho odtoku vody z\u00a0n\u00e1dr\u017ee.<\/p>\n

Aplikace<\/h2>\n

Model byl aplikov\u00e1n na fiktivn\u00ed n\u00e1dr\u017e na \u0159ece Svitav\u011b (obr. 2<\/em>, m\u011brn\u00fd profil B\u00edlovice nad Svitavou). M\u00edsto aplikace bylo vybr\u00e1no pro dostupnost dat. N\u00e1dr\u017e byla navr\u017eena tak, aby do\u0161lo k\u00a0v\u00fdrazn\u00fdm poruch\u00e1m b\u011bhem tr\u00e9novac\u00edho obdob\u00ed (d\u00e9lka 15\u00a0let). N\u00e1dr\u017e byla \u0159\u00edzena adaptivn\u011b s\u00a0\u010dasov\u00fdm krokem jeden m\u011bs\u00edc. \u0158\u00eddic\u00ed odtok byl nastaven na hodnotu 4,25\u2008m3<\/sup>\/s\u00a0a\u00a0maxim\u00e1ln\u00ed objem vody v\u00a0n\u00e1dr\u017ei byl stanoven na 19,76\u2008m3<\/sup>\/s\u00a0(objem je uveden v\u00a0pom\u011brn\u00fdch jednotk\u00e1ch jako pod\u00edl objemu za jednotku \u010dasu, pokud by z\u00a0n\u00e1dr\u017ee byla vypou\u0161t\u011bna hodnota pom\u011brn\u00e9ho objemu po dobu 1 pr\u016fm\u011brn\u00e9ho m\u011bs\u00edce = 2 630 000\u2008s, byl by cel\u00fd z\u00e1sobn\u00ed objem n\u00e1dr\u017ee vypr\u00e1zdn\u011bn). Maxim\u00e1ln\u00ed objem vody v\u00a0n\u00e1dr\u017ei a\u00a0\u0159\u00eddic\u00ed odtok byly zvoleny tak, aby doch\u00e1zelo v\u00a0testovac\u00edm obdob\u00ed k\u00a0v\u00fdrazn\u00fdm poruch\u00e1m (nedod\u00e1vce vody). Model provedl \u0159\u00edzen\u00ed s\u00a0pravd\u011bpodobnost\u00ed odtoku 99, 95, a5, 5, 0,01\u2009% pro d\u00e9lky p\u0159edpov\u011bd\u00ed 1 a\u017e 12\u00a0m\u011bs\u00edc\u016f. P\u0159edpov\u011bdi pr\u016fm\u011brn\u00fdch m\u011bs\u00ed\u010dn\u00edch p\u0159\u00edtok\u016f byly z\u00edsk\u00e1ny z\u00a0modifikovan\u00e9ho zon\u00e1ln\u00edho line\u00e1rn\u00edho autoregresivn\u00edho rekurentn\u00edho modelu [5].<\/p>\n\"Kozel-2\"<\/a>\n

Obr. 2. M\u011brn\u00fd profil B\u00edlovice nad Svitavou
\nFig. 2. Measured profile B\u00edlovice nad Svitavou<\/h6>\n

Vyhodnocen\u00ed<\/h2>\n

Hlavn\u00edm krit\u00e9riem \u00fasp\u011b\u0161nosti \u0159\u00edzen\u00ed byla druh\u00e1 mocnina chyby mezi v\u00fdstupy z\u00a0modelu pou\u017e\u00edvaj\u00edc\u00ed p\u0159edpov\u011bzen\u00e9 p\u0159\u00edtoky a\u00a0\u0159\u00edzen\u00edm, kter\u00e9 bylo z\u00edsk\u00e1no z\u00a0klasick\u00e9ho modelu za pou\u017eit\u00ed 100% p\u0159edpov\u011bdi (re\u00e1ln\u00e1 pr\u016ftokov\u00e1 \u0159ada) [6]. Po\u010det zp\u011btn\u011b uva\u017eovan\u00fdch krok\u016f siln\u011b ovliv\u0148oval pr\u016fb\u011bh \u0159\u00edzen\u00ed a\u00a0pravd\u011bpodobnost, pro kterou byly dosa\u017eeny nejlep\u0161\u00ed v\u00fdsledky (obr. 3<\/em>). Za nejlep\u0161\u00ed pr\u016fb\u011bh \u0159\u00edzen\u00ed bylo pova\u017eov\u00e1no nejvyrovnan\u011bj\u0161\u00ed \u0159\u00edzen\u00ed s\u00a0m\u011blk\u00fdmi poruchami. Uveden\u00e9 obr\u00e1zky graf\u016f maj\u00ed toto\u017en\u00e9 osy, kdy na svisl\u00e9 ose je v\u017edy \u0159\u00edzen\u00fd odtok vody z\u00a0n\u00e1dr\u017ee pro vybranou pravd\u011bpodobnost odtoku vody z\u00a0n\u00e1dr\u017ee a\u00a0na vodorovn\u00e9 ose je \u010das v\u00a0m\u011bs\u00edc\u00edch. V\u00a0legend\u011b \u010d\u00edslo uveden\u00e9 za P<\/em> (pravd\u011bpodobnost) zna\u010d\u00ed uva\u017eovanou pravd\u011bpodobnost \u0159\u00edzen\u00e9ho odtoku vody z\u00a0n\u00e1dr\u017ee.<\/p>\n\"Kozel-3\"<\/a>\n

Obr. 3. V\u00fdsledky pro r\u016fzn\u00fd po\u010det p\u0159edchoz\u00edch krok\u016f \u0159\u00edzen\u00ed pou\u017eit\u00fdch pro pr\u016fm\u011brov\u00e1n\u00ed
\nFig. 3. Result for different numbers of backward steps of management used for averaging<\/h6>\n

Druh\u00fdm v\u00fdznamn\u00fdm faktorem, kter\u00fd ovlivn\u00ed celkov\u00fd pr\u016fb\u011bh \u0159\u00edzen\u00ed, je d\u00e9lka p\u0159edpov\u011bdi. Z\u00a0v\u00fdsledk\u016f bylo zji\u0161t\u011bno (obr. 4<\/em>), \u017ee nejlep\u0161\u00edch v\u00fdsledk\u016f bylo dosa\u017eeno p\u0159i d\u00e9lce p\u0159edpov\u011bdi \u0161est m\u011bs\u00edc\u016f a\u00a0pou\u017eit\u00ed \u010dty\u0159 krok\u016f \u0159\u00edzen\u00ed zp\u011btn\u011b pro pr\u016fm\u011brov\u00e1n\u00ed. P\u0159i pou\u017eit\u00ed krat\u0161\u00ed d\u00e9lky p\u0159edpov\u011bdi ne\u017e \u0161est m\u011bs\u00edc\u016f je velmi zmen\u0161en po\u010det pravd\u011bpodobnost\u00ed, pro kter\u00e9 nedojde k\u00a0\u00fapln\u00e9mu vy\u010derp\u00e1n\u00ed z\u00e1sobn\u00edho objemu (P99, P95), a\u00a0um\u011bl\u00e9 poruchy jsou hlub\u0161\u00ed ne\u017e pro del\u0161\u00ed p\u0159edpov\u011b\u010f. Pou\u017eit\u00ed del\u0161\u00ed p\u0159edpov\u011bdi ne\u017e \u0161est m\u011bs\u00edc\u016f nep\u0159ineslo v\u00fdrazn\u00e9 zlep\u0161en\u00ed. V\u00fdsledky pro d\u00e9lku p\u0159edpov\u011bdi osmi a\u00a0\u0161esti m\u011bs\u00edc\u016f jsou t\u00e9m\u011b\u0159 toto\u017en\u00e9, uveden\u00e9 tvrzen\u00ed neplat\u00ed pro pot\u0159ebnou d\u00e9lku v\u00fdpo\u010dty.<\/p>\n\"Kozel-4\"<\/a>\n

Obr. 4. Vliv d\u00e9lky p\u0159edpov\u011bdi na pr\u016fb\u011bh \u0159\u00edzen\u00ed
\nFig. 4. Influnce lenght of forecast on course of management<\/h6>\n

B\u011bhem \u0159\u00edzen\u00e9ho obdob\u00ed model vytv\u00e1\u0159el um\u011bl\u00e9 poruchy, av\u0161ak pr\u016fb\u011bh \u0159\u00edzen\u00ed byl velmi dobr\u00fd i\u00a0p\u0159i srovn\u00e1n\u00ed s\u00a0klasick\u00fdm modelem (100% p\u0159edpov\u011b\u010f) (obr. 5<\/em>).<\/p>\n\"Kozel-5\"<\/a>\n

Obr. 5. Porovn\u00e1n\u00ed stochastick\u00e9ho modelu a\u00a0deterministick\u00e9ho modelu
\nFig. 5. Comparison between results of stochastic model and results of deterministic model<\/h6>\n

Z\u00e1v\u011br<\/h2>\n

Z\u00a0v\u00fd\u0161e uveden\u00e9ho textu vypl\u00fdv\u00e1, \u017ee v\u00fdsledn\u00e9 \u0159\u00edzen\u00ed je z\u00e1visl\u00e9 na po\u017eadovan\u00e9 pravd\u011bpodobnosti, s\u00a0n\u00ed\u017e by se m\u011bly p\u0159\u00edtoky vyskytnout a\u00a0s\u00a0nimi i\u00a0po\u017eadovan\u00fd odtok vody z\u00a0n\u00e1dr\u017ee. Nelze opomenout, \u017ee v\u00fdsledn\u00e9 \u0159\u00edzen\u00ed je z\u00e1visl\u00e9 na celkov\u00e9m po\u010dtu i\u00a0d\u00e9lce p\u0159edpov\u011bd\u00ed a\u00a0po\u010dtu krok\u016f uva\u017eovan\u00fdch p\u0159i \u0159\u00edzen\u00ed zp\u011btn\u011b. P\u0159i vy\u0161\u0161\u00edm po\u010dtu p\u0159edpov\u011bd\u00ed doch\u00e1z\u00ed ke zlep\u0161en\u00ed \u0159\u00edzen\u00ed. Vyhodnocen\u00ed stochastick\u00e9ho \u0159\u00edzen\u00ed vy\u017eadovalo velk\u00e9 mno\u017estv\u00ed strojov\u00e9ho \u010dasu a\u00a0v\u00fdpo\u010dty na clusterech. Srovnateln\u00fd v\u00fdpo\u010det (150\u00a0m\u011bs\u00edc\u016f, d\u00e9lka p\u0159edpov\u011bdi 8 a\u00a0po\u010det p\u0159edpov\u011bd\u00ed 500, 10 pravd\u011bpodobnost\u00ed) trval na clusteru 10\u00a0hodin, na b\u011b\u017en\u00e9m stoln\u00edm PC bylo pot\u0159eba 7\u00a0dn\u00ed. V\u00fdsledky ukazuj\u00ed, \u017ee lze z\u00e1sobn\u00ed funkci n\u00e1dr\u017ee \u0159\u00eddit stochasticky pomoc\u00ed metody DE a\u00a0stochastick\u00e9ho modifikovan\u00e9ho zon\u00e1ln\u00edho p\u0159edpov\u011bdn\u00edho modelu.<\/p>\n

Pod\u011bkov\u00e1n\u00ed<\/h3>\n

\u010cl\u00e1nek vznikl za podpory projektu \u0158\u00edzen\u00ed z\u00e1sobn\u00ed funkce n\u00e1dr\u017e\u00ed p\u0159i uva\u017eov\u00e1n\u00ed nejistot hydrologick\u00fdch vstup\u016f s\u00a0pou\u017eit\u00edm metod um\u011bl\u00e9 inteligence s\u00a0podporou stochastick\u00fdch p\u0159edpov\u011bdn\u00edch model\u016f.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

Described models are used random forecasting period of flow line with different length. The length is shorter than 1 year. Forecasting period of flow line is transformed to line of managing discharges with same length as forecast. <\/p>\n","protected":false},"author":8,"featured_media":3458,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[86],"tags":[831,656,829,830],"coauthors":[805,806],"class_list":["post-3504","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-hydraulics-hydrology-and-hydrogeology","tag-evolution-algorithm","tag-open-water-reservoir","tag-stochastic","tag-storage-function"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/3504","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/comments?post=3504"}],"version-history":[{"count":2,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/3504\/revisions"}],"predecessor-version":[{"id":30395,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/posts\/3504\/revisions\/30395"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/media\/3458"}],"wp:attachment":[{"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/media?parent=3504"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/categories?post=3504"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/tags?post=3504"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/www.vtei.cz\/en\/wp-json\/wp\/v2\/coauthors?post=3504"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}