ඕЪˠ̍ംᇊᄃछۢᙊ̝ംᇊݭͪऱፆүր
Integrating AI with Expert Knowledge to Build
Intelligent Reservoir Operation System
઼ϲέ៉̂ጯϠۏᒖဩր ̍ጯրି
ૺ!!ౢ!
Fi-John Chang
઼ϲέ៉̂ጯϠۏᒖဩր ̍ጯր౾̀ࡁտϠૺ!ฮ!ನ!
Ya-Ting Chang
୶ѯ̂ጯͪྤ̈́ᒖဩ ̍ጯրӄநିૺ!ᚊ!ࡌ!
Li-Chiu Chang
ĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝၡ! ! ࢋ
ࢬ၆έ៉гડͪྤॡ۩̶Ҷ̙Ӯ̈́͟ৈ̙֖ඈયᗟĂтңдщБ୧І˭ซ ҖͪऱፆүֹጐΞਕ႕֖Чᇾ۞Ăͽචϡͪྤăჯᒖဩϖᜈّߏ༊݈ࢵࢋ ኝᗟĄώࡁտͽາ᎖۞ˠ̍ംᇊ࠹ᙯநኢĂ֭ඕЪனҖఢቢፆү۞छۢᙊ೩ ംᇊݭͪऱፆүඉரĂͽϮܝͪऱ࿅Ν36 ѐ̝ͪ͛ېڶࠎּĂซҖ၁ચሀᑢീྏć ࢵАӀϡ็ႊზڱ(GA)ವՐ።Ϋ߹ณ̝ͪऱָٸͪณ።Ăͽਬүࠎአዋّ შྮሀቘଯኢր(ANFIS)̝ቚᇹώᄃᇾ۞Ąࠎᆧΐրፆүఢऱ̝Ԇፋّᄃ ЪڱّĂ˜ࡁᛉͪऱፆүఢቢᄃሀቘఢऱ̝ม۞ᖼೱ͞ёᄃ፟טĂፆүఢቢ ܑ̝ٙᄊٸᇾᖼೱࠎఢ(if-then)ԛёĂޙၹሀቘఢۢᙊऱĂјΑ۞ͪ ऱ็۞ፆүඉரᄃംᇊݭፆүሀёซҖඕЪĂᖣϤΐˢ็ፆү͞ё۞छۢ ᙊֹրՀĺംᇊĻгநྤफ़ᄃҿᕝྤੈĂซ҃ѣड़гଠטͪऱͪҜᄃٸ ߹ณĂ೩ֻͪऱგநಏҜٺᄊͪӀϡྻᖼॡѣٙણ҂ֶ̈́ፂĂീྏඕڍពϯώࡁ տٙ൴ण۞ሀёྵ็ఢቢፆү͞ёдЧีᑭീᇾ˯࠰ѣ಼̂۞ԼචĂϺОᙋ ˞ሀё۞Ъநّᄃዋ̷ّĄ ᙯᔣෟĈͪऱፆүĂˠ̍ംᇊĂ็ႊზڱĂሀቘఢऱĂአዋّშྮሀቘଯኢր ĄABSTRACT
Resulting from the continuous increase in water demand and uneven water distribution both on time and space, the efforts of pursuing integrated optimal water resource management become critical. In this study, we propose a novel intelligent control methodology that includes the genetic algorithm (GA), fuzzy rule base (FRB), and the adaptive network-based fuzzy inference system (ANFIS) to enhance the
ྺຽ̍ጯಡ! ௐ50 סௐ 4 ഇ Journal of Chinese Agricultural Engineering ̚රϔ઼93 ѐ 12 ͡ۍ Vol. 50, No. 4, December 2004
Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ
efficiency of reservoir operation. The Shihmen reservoir in north Taiwan is used as a case study, and its last thirty-six years hydrological data are used to train and/or verify the models’ performance. GA and FRB are used to extract the knowledge based on the historical inflow data with a design objective function and the traditional rule curve operating strategy, respectively. The ANFIS is then used to implement the knowledge, to create the fuzzy inference system, and then to estimate the optimal reservoir operation. The practicability and effectiveness of the proposed approach is tested on the operation of the Shihmen reservoir. The results show that the ANFIS models built on different types of knowledge have better performance than the traditional M-5 rule curves in reservoir operation. Moreover, we demonstrate that the ANFIS model can be more intelligent for reservoir operation if more information (or knowledge) is involved.
Keywords: Reservoir operation, Artificial intelligent, Genetic algorithm, Adaptive
network-based fuzzy inference system, Fuzzy rule base.
ĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝĝ
˘ă݈! ֏
ͪྤ่̙ߏჯˠᙷϠх̙ٙΞٕ۞ ࢦࢋྤĂՀߏགྷᑻ൴ण۞ᙯᔣࢋ৵Ą҃έ៉Я ͇ᒖဩࢨטĂͪྤצঈ෪ͪ͛ăቛޘăгԛ ୧ІඈኜкЯ৵ᇆᜩĂѣޘ̙ቁؠّĂ่̙ ؞༼ّ̝ܥณᖳߜ̶Ҷ̙Ӯ̹ĂೇЯгԛ̝г๕ ੲथĂͪྤᄊ᎕̙ٽĂౄјͪྤྻϡ͟ৈӧ ᙱĄ ѝഇέ៉гડ̝ͪྤӀϡ˜ߏͽྺຽϡ ͪࠎĂז˞ϔ઼ 60 ѐޢഇĂЯۤົඕၹ ԼតĂੵ˞၆ͪᘦؠّࢋՐྵҲ̝ྺຽϡͪณ ϒుѐ˭ࢫγĂϔϠᄃ̍ຽϡͪᅮՐӮ֝ిј ܜĂ҃າͪฟ൴פ͟ৈӧᙱĂफ˯ͪྤ ̏ࢬᓜĶͪķᓜࠧᕇĂ่̙၆ٺ઼छϏֽགྷᑻ ൴णԛјᐚĂϺ၆έ៉ϖᜈ൴णԛјᅪᘣĂܕ ೀѐ۞ͪਣયᗟߊࠎځᙋĄ дгܑͪӀϡ͞ࢬĂͪऱࠎ˘ࢦࢋᄊͪନ ߉Ăдߜͪ؞(ګ̌߹ณҲĂ͔̙֖ͪ)Ξ೩ֻአ ᖳᑻߜ۞Αड़Ăଘ։рឥӬк̏ฟ൴ĂЯѩܕѐ ֽ ͪ ྤ ૄ ώ ඉ ர ̏ ࣒ ϒ ࠎ ͽ አ ޘ გ ந ࠎ ᐹ АĄϤٺ̙Т۞ͪऱፆү͞ڱ၆ͪऱ۞ፋវֹϡ ड़தົౄјࢦ̂ᇆᜩĂፆү̙։ົࢫҲͪऱֻ ͪड़தĂౄјͪྤڱ·̶Ӏϡͷٽጱ˭ഫ ̝ͪன෪ĄЯѩĂࡁտтңдщБ୧І˭၆ͪ ऱซҖፆүֹጐΞਕ႕֖Чڇચᇾ۞Ăͷѣड़ ྻϡனѣͪऱྤֹͪऱϖᜈ൴णགྷᒉĂ၁ࠎ˘ ࢦࢋࡁտኝᗟĄ ࠎՐዋϡٺ࿅ΝϏഅٕޝ͌൴Ϡ̝ໂბͪ ͛ன෪Ă̈́ЯᑕЧᇾ۞ϡͪᅮՐ̏Яۤົགྷᑻඕ ၹតዏ҃Լត̝ଐ๕Ăώ͛ؼᜈ࿅Ν̝ࡁտ(ૺᚊ ࡌăૺౢ, 1999)೩˘҂ᇋᆸࢬྵ̂ͷྵѣր ̶̝ژ͞ёĂӈӀϡˠ̍ംᇊநኢ̝ᇅّඕЪ ็ఢቢፆү̝ᐹᕇޙၹͪऱፆүրĂֹր ౯֖ૉ۞ംᇊซҖҿᕝᄃՙඉĂ֭ซ˘Վֹͪ ऱგநಏҜٺᄊͪӀϡྻᖼॡѣٙણ҂ֶ̈́ፂĄ րߛၹтဦ1 ٙϯĂࠎޙϲംᇊݭͪऱፆ үրĂࢵАᅮࢋ።Ϋָٸ߹።үࠎሀё̝ ቚྤफ़Ă҃၁ᅫͪऱፆү֭˘इĺָĻ ̝ٸ߹።ĂЯѩࡁտࢵАޙϲͪऱፆү̝ϫᇾ בᇴᄃࢨט୧ІĂГॲፂͪऱ።Ϋˢ߹ԔЕͽ ็ႊზڱᐹᏴ˘րָྋĂүࠎͪऱ̝። Ϋָٸ߹።Ăѩ።Ξүࠎޢᜈംᇊݭଠ טր̝Ꮾˢ—Ꮾቚྤफ़Ąംᇊݭͪऱፆү րͽአዋّშྮሀቘଯኢր(ANFIS)ࠎߛ ၹĂΒӣ˞ˬఢऱĂ̶Ҿࠎ(1)Ϥ GA ٙᐹᏴ ۞ָٸ߹።ĂᏴϡЪዋ̝ሀቘჸᙷ͞ڱ Տ˘ඊϤᏮˢШณᄃᏮШณЪј۞ྤफ़ΐͽ ̶ᙷ֭ྻϡٺሀቘIf-then ఢĂޙၹሀቘఢ ऱGAć(2)ΐˢ็ፆү͞ё۞छۢᙊĂӈ ͪऱፆүఢቢᖼೱࠎሀቘఢۢᙊऱ FRBć(3) ͽ˯˟ఢऱ̝ඕЪ(GA & FRB)Ąώࡁտࢦᕇ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ Ĝ& GA ဦ1 ംᇊݭͪऱፆүրߛၹဦ ӈߏࡁᛉͪऱፆүఢቢᄃሀቘఢۢᙊऱ̝ม ۞ᖼೱ፟טĂ็ఢቢፆү̝छۢᙊᄃ࿅Ν ࡁտٙޙϲ̝ംᇊݭͪऱፆүրඕЪĂሀё ่̙Βӣ።Ϋྤफ़ٙᔳӣ۞ྤੈĂТॡ˵౯˞ छፆү̝ۢᙊᄃགྷរĂֹրΞംᇊгଠטͪ ऱͪҜᄃٸ߹ณĄ
˟ă͛ᚥаᜪ
Ϥٺۤົඕၹ̝ԼតĂ઼ˠϡͪᅮՐ͟ᆧĂ ֹͪྤ̝አ੨ăྻϡֶֽᏥͪऱ۞ᐼх ᄃአ༼Ąͪऱፆүඉர̝ࢎؠ˘ਠѣሀᑢڱ̈́ᐹ Ᏼ ڱ ˟ Ą ็ ͪ ऱ ፆ ү ఢ ቢ(operating rule curves)ćHEC-3ăHEC-5 ሀё(Ϥ Hydrological Engineering Center ̶Ҿٺ 1971 ᄃ 1979 ѐ൴ण) ࠰ࠎሀᑢሀё̝ٙඕڍćѦုᅞඈ(2000)˵അ ଣሀᑢڱдͪऱፆүఢቢ̝ᑕϡĄፆүఢቢ˜ ߏ ͪ ऱ д న ࢍ ఢ ထ ล ߱ ӈ ॲ ፂ Ԇ ̍ ॡ ۞ ᄊ ͪ ณăֻ(ϡ)ͪณĂ֭੨ЪะͪડՏџซͪณ̝ঈ ෪ྤफ़ඈซҖሀᑢĂГᖣϤྏᄱڱ(try & error)ႊ ზĂϤкΞਕ̝࣏Ᏼఢቢ̚߄Ᏼ˘ߊ႕֖ Чᇾ۞ϡͪ˫ٽٺፆү۰ֹٙϡ̝ፆүఢቢĄ έ៉гડனѣͪऱдፆү˯кଳྻᖼఢቢ ͞ёĂӈͪऱᄊͪ۩ม̶ࠎࡶ̒࣎ડાтĶ˯ ࢨķăĶ˭ࢨķăĶᚑࢦ˭ࢨķ̈́ĶӑͪҜķඈĂ ̶Чᄊͪડા̝ᄊͪณࠧࢨӈࠎፆүఢቢĂ఼ ૱Чડม̂̈ᐌ؞༼ត̼҃ѣٙमளĂፆүఢ ̚ ֭ ט ؠ Ч ॡ ഇ ̙ Т ᄊ ͪ ณ ٙ ၆ ᑕ ̝ ٸ ͪ ࣧ ĄϤٺЧͪऱ̝পّᄃ˭ഫϡͪᅮՐ̙ТĂੵ ӑͪҜؠγĂЧџͪҜ̝ፆүఢቢᇾ࠰ѣ ˘ؠ۞ఢؠĂͪऱ̝ͪҜଠטᅮॲፂፆүఢ ቢซҖგநĂͽቁܲͪऱ۞ᇅّ͚೯ᄃአᖳᑻ ߜ̝ΑਕĄ ӀϡᐹᏴڱޙϲͪऱፆүሀёĂืͽᇴந ͞ёୃϫᇾבᇴᄃրྻүԔĂӀϡՐྋ ԫఙವՐ႕֖ࢨט୧І˭Ăֹϫᇾבᇴܑனָ ۞˘ՙඉតᇴĄܕѐֽϤٺཝጸटณᄃి ޘ֝ి೩چĂྻϡᐹᏴڱซҖͪऱ۞ָఢထ͟ ᔌΞҖĂЧႊზڱ၆ኑᗔ۞ఢထયᗟϺѣ̙ ۞ᐹᏴਕ˧Ą ᐹᏴڱᑕϡٺͪऱፆүඉர̝࠹ᙯࡁտ ѣĈోॎăૺڠ(1984)ćKelman et al.(1990)ć ˜ᯂ(1997)ćThomas et al.(1997)ćPerera & Codner(1998)ćᏂܛੑăૺ։ϒ(1998)Ԕதָ ፆүሀёᑕϡٺкϫᇾͪऱրćౘᙶт(2001) ͽሀᑢੜͫڱᐹᏴ̂ϥ˭ഫָአᄊͪѰ̝ ፆүఢćषѐăเୂം(2001)ඕЪሀᑢੜͫ ڱᄃͪऱፆүሀᑢሀёĂՙؠ͟͡ሔָ̝ፆү ఢቢćChang et al.(2002)ćChandramouli et al.(2002) ඈˠજၗఢထᑕϡٺкϫᇾٕкͪऱր ָፆүඉர̝ࡁᛉćܘߒ(2003)ߏͽЪё ็ੜͫႊზڱຩವͪऱܜഇፆүָ̝ᒉྻ ඉரĄ ࠎࡁᛉͪऱӈॡፆүඉரĂૺᚊࡌăૺౢ (1999)Ᏼϡܕѐֽᇃࠎ൴ण۞็ႊზڱăሀቘ ଯኢ̈́ᙷৠགྷშྮඈംᇊݭଠטநኢĂޙϲ˘አ ዋّሀቘଠטրซҖͪऱፆүĄംᇊݭଠטந ኢࢋΒӣˠ̍ംᇊăछրăሀቘநኢă ็ႊზڱᄃᙷৠགྷშྮඈԫఙĂΞሀᑢˠᙷጯ ௫ăዋᑕăаຐඈኜкਕ˧Ăྋՙሀё̙ቁؠր ăܧቢّᄃॡតրඈ็͞ڱ̙ٽྋՙ̝ય ᗟĂϫ݈̐јΑᑕϡٺଠטЧё྿ᄃ፟ၹĂт ፟ጡˠăˠዼዺ֘ዃăଥྻăୗᄃЧёࢳҖ ጡඈ(Davies & Watton, 1995ćLin and Su, 2000ć Becerikli et al., 2003)Ąˬăநኢໄ
дޙၹംᇊݭଠטր̝݈ĂࢵАᅮ၆ր ۞ۢᙊٕྤੈѣٙ˞ྋĂֱۢᙊٕྤੈ۞ܑன
͞ёΞͽߏఢݭёĂځቁгܑனր၆ٺᏮ ˢࣃ۞ͅᑕଐԛĂٕᖣϤ።Ϋྤफ़۞ќะĂଂ̚ ᒔᏮˢ—Ꮾม̝ᙯܼĄϤٺംᇊݭͪऱፆү րᅮᖣѩྤੈซҖሀё̝ߛၹᄃቚĂҭ።Ϋ ྤफ़̚ͻѩੈिĂЯѩืՐפ˘நຐ۞ͪऱ ٸ߹።ྤफ़ͽ౯ϏֽሀёቚॡഇֹϡĄޞሀ ёߛၹԆјĂ˘όϏֽ൴ϠᙷҬ۞ͪ͛ېڶĂം ᇊݭͪऱፆүրӈΞણ҂࿅Ν۞ۢᙊүҿᕝĄ ͽ˭ᖎಏ̬ࡁտଳϡ̝็ႊზڱăሀቘ ଯኢրͽ̈́አዋّშྮሀቘଯኢրඈംᇊ ݭଠטநኢ۞ొ̶ૄώໄهćࡶԓ୕ᒔՀஎˢ ۞˞ྋĂΞણ҂ૺᚊࡌăૺౢ(1999)̝үĄ 3.1 ็ႊზڱ(genetic algorithm) ็ႊზڱܼJohn Holland ٺ 1975 ѐ൴ܑ ኢ “Adaptation in Natural and Artificial Systems”ٙ̚೩Ă៍هٺ྿Ⴌ͛ซ̼ኢ̚ ĺۏᚮ͇ፄĂዋ۰ϠхĻ̝ጯᄲĂૻአͽૄЯ ആ࿅ـ̝ྻზ̮ᇴфĂયᗟᖼೱࠎҋࠧႊ̼ ԔĂ֭ٺՐྋָ̼યᗟ࿅̚ሀᑢĶۏႊ ̼ķ۞ҖࠎĂЯ҃൴णј˘Бાຩವ(Global Search)۞ႊზڱĂጯ௫ր˜ߏሀᑢཏะ ็ᄃዋᑕ۰Ϡх̝࿅ֽᆧซඕڍܑனĂҌ ̫̏јΑྋՙ˞̂ొ̶็ྋژᄃᇴࣃ۞ָ ̼ԫఙٙᙱͽՐྋ̝בᇴָ̼યᗟĄ ็ႊზڱፋ࣎ᐹᏴ߹̂ࠎĈ็࿅ ̚ཏะצטٺዎᒖဩĂֹዋᑕ˧ָ۞јࣶజᏴ ࠎ੨၆ᄃኑᄦ̝ᏐĂЯѩܑனྵр఼̝̄૱ ߏϤྵᐹս̝Ꮠᗕ͞็ֽ҃Ăז˞ௐ˟ዋ ᑕ։р۞јࣶ˫జᏴֽซҖ੨၆ăኑᄦĂᜈซ ҖᚮۋёೈᒖĂֹܑனम۰ዎז՛Ăܑ னᐹ։۰யϠՀָ̝ޢĂтѩᓄࢉĂᇴ ޢٙх߿̝ཏะӈࠎዋٺᒖဩϠх۰Ą ܕѐֽధкࡁտ็ႊზڱ၆ٺኑᗔ ۞ఢထયᗟѣ։р۞Րྋਕ˧Ă࠹ᙯࡁտѣĈ ౢ(1994)ᑕϡٺ̼ͪ͛ጯրણᇴ̝ᐹ Ᏼćเॎཐ(1995)ͽ็ႊზڱᐹᏴഅ͛ͪऱџ ፆүఢቢ֭ଣͪऱࢲᐍćధ͌༉(2001)ᑕϡ ็ ႊ ზ ڱ ᐹ Ᏼ ͪ ऱ ፆ ү ఢ ထ ય ᗟ ۞ ଠ ט ᕇ ሀ ёĄChang ඈ(2004)ྻϡٺϮܝͪऱፆүఢቢ̝ ဦ2 ሀቘଯኢր̝ૄώߛၹ ᐹᏴć઼γϺѣ̙͌ࡁտᑕϡ็ႊზڱֽྋՙ Чͪྤયᗟ(Wang, 1991ćMantawy et al., 1999ćCai et al., 2001)Ą ͪऱፆү۞ఢထયᗟϤٺតᇴྵкͷϫᇾ בᇴᄃࢨטёኑᗔĂ็۞ቢّٕܧቢّఢထ̙ ٽՐྋĄϤٺώࡁտᑕϡ၁ּ̝።Ϋྤफ़่ѣ። ѐͪऱˢ߹ณăࢍ൪ᅮͪณăͪऱፆүఢቢඈĂ ҃ͻָٸ߹ณྤफ़үࠎޢᜈ ANFIS ሀё̝ ቚྤफ़Ă߇Ӏϡ็ႊზڱᐹᏴ።ѐ࠹ᙯ۞ ͪऱटณត̼ྤफ़ᄃָٸ߹።Ą
3.2 ሀቘଯኢր(Fuzzy Inference System)
ሀቘଯኢր˫Ⴭࠎሀቘఢऱրăሀቘ ଠטٕߏሀቘᓑຐጸ(FAM)Ăߛၹтဦ 2 ٙ ϯĂΒ߁Ĉሀቘ̼(fuzzifier)ăሀቘఢ(fuzzy rules)ăᔴᛳבᇴ(membership function)ྤफ़ऱă ଯኢ͔ᑜ(inference engine)ᄃྋሀቘ̼(defuzzifier) ඈ̣̂ొ̶Ăϫ݈̏јΑгᑕϡٺҋજଠטăྤ फ̶़ᙷăՙඉ̶ژăछրඈ̙ТᅳાĄдͪ ྤ͞ࢬ̝ᑕϡѣĈૺౢඈ(1993)ޙϲሀቘଯ ኢሀёĂᑕϡٺͪ͛ր̝ࡁտćୖॢᅛඈ (2000)ซ˘ՎඕЪᙷৠགྷშྮĂͽኑЪႊზᙷ ৠགྷęሀቘଯኢሀёซҖ߸ͪീ̝ࡁտĂ࠰ѣ ࠹༊̙۞ࡁտјڍĄ ሀቘଯኢ̝ՎូࠎĈ(1)ሀቘ̼ĈͧྵᏮˢត ᇴ݈೩ี(premise)ొ̶̝ᔴᛳבᇴĂͽᒔՏ࣎ ᄬຍ̝ᔴᛳޘć(2)ඕЪՏ࣎ఢ̝݈೩ีొ̶۞ ᔴᛳޘĂͽזfiring strength(ӈᝋࢦࣃ)ć(3)ֶ ፂ firing strength ய Ϡ Տ ࣎ ఢ ̝ ඕ ኢ ี (consequent)ણᇴć(4)ྋሀቘ̼ĈϤඕኢีણᇴய Ϡ˘ځቁᏮࣃĄ 3.3 አዋّშྮሀቘଯኢր(adaptive network-
based fuzzy inference system) 3.3.1 አዋّშྮ
አዋّშྮ˜ߏ˘кᆸ݈㒝ёშྮĂშྮඕ ၹ̚Βӣ˞༼ᕇᄃ༼ᕇม̝ాඕĂͷЧ༼ᕇבᇴ ࠹ҬăአዋّĂ҃༼ᕇᏮ˜ߏֶፂણᇴٙ ՙؠĂЯѩშྮጯ௫ڱߏдአፋણᇴֹᄱमࢫ ҲĄ୬ޙϲሀቘሀёĂࢵАᅮࢎᏮˢᄃᏮត ᇴ۞ᙷăᇴณᄃᙷݭĈ (1) ᏮˢតᇴĈϤٺᇆᜩͪऱٸͪЯ৵ிкĂ ڱۡᛇҿᕝٕ࿅ր̶ژԱᄊͪ ณăˢ߹ณăᅮͪณᄃٸ߹ณม۞ᙯܼĂ ЯѩᅮࢋҋᏮˢᄃᏮྤफ़̚ĂӀϡྏᄱ ڱ˯តᇴͽ̙Т۞Ъˢሀё̚Ă Աቚᄱमࣃ̈۞Ъ͞ёĂүࠎሀ ё۞ᏮˢតᇴĂͽቁܲቚ̝ሀёΞѣ ड़۞ೡᇴፂ۞পّĄ (2) ᏮតᇴĈᏮតᇴΪѣ˘࣎Ăӈͪऱٸ ͪณĂ҃ޙၹ۞ሀёଳϡ˘ลsugeno ሀቘ ଯኢሀёĂ߇ᏮតᇴࠎᏮˢតᇴ۞ቢّ בᇴĄ ᏮˢᄃᏮតᇴՙؠޢĂనؠЧ࣎តᇴ۞ᔴ ᛳבᇴᙷݭᄃᇴณĂӀϡቚྤफ़ֽአፋЧีણ ᇴĂ࠽ѩሀቘሀё۞ᏮඕڍਕՀࠎЪࣧؕᇴ ፂĄ 3.3.2 አዋّშྮሀቘଯኢր አዋّშྮሀቘଯኢր(Jang, 1993)ߏͽ ሀቘଯኢրࠎშྮሀёૄᖂĂ֭ඕЪৠགྷშྮ ҋԧᖐ۞পّߛၹ҃јĄሀቘଯኢրᖣϤሀ ቘ If-then ఢ ၆ ٺ ˠ ᙷ ۢ ᙊ ᄃ ଯ ኢ ࿅ (reasoning processes)ซҖؠّೡᄃ̶ژĂҭߏ ݒͻቁ۞ؠณ̶ژᄃᇴࣃ७ϒć҃ᙷৠགྷშ ྮᔵڱநؠّ۞ۢᙊᄃទᏭଯኢ࿅Ăݒ ѣໂָ۞ҋԧጯ௫ᄃᖐਕ˧Ăૻ̂۞አፋਕ ˧ϒΞϡֽүሀቘր۞ඕၹᄃણᇴ̝አፋĄЯ ѩĂANFIS ඕЪ˞˟ႊზڱĂΞ·̶൴೭ሀё ၆ ٺ ր ̙ ቁ ؠ ّ(uncertainty) ᄃ ̙ ჟ ቁ ّ (imprecisely)۞நਕ˧Ă࿅ ANFIS ጯ௫ᄃҋ ԧአዋซ҃ՐણᇴָྋĄ ͽ˭ͽ˟࣎Ꮾˢࣃă˘࣎Ꮾࣃࠎּăᄲځ րߛၹᄃ˟ล߱ጯ௫Ăߛၹтဦ3 ٙϯĄ ௐ˘ᆸ ᏮˢᆸĈᏮˢតᇴߍडҌሀቘะ ЪĂͽనؠ̝ᔴᛳבᇴҤზᔴᛳޘĂనଳϡ x y x y x y A1 A2 B1 B2 Π Π N N Σ ဦ3 ANFIS ߛၹဦ S ݭ(Sigmoidally-shaped)ᔴᛳבᇴĂт˭ёٙϯĈ 4 , 3 ) ( 2 , 1 ) ( 2 , 1 , 1 = = = = − y for i O i for x O i i B i A i µ µ ... (1) ̚O1ࠎᏮˢࣃ࠹၆ٺሀቘะЪ̝ᔴᛳבᇴĂ ) ( 1 1 i i i a x c A e− − + = µ Ă ( ) 1 1 2 i i i a y c B e− − + = − µ Ă {a ,i ci } ࠎ ሀ ቘ ᔴ ᛳ ב ᇴ ۞ ણ ᇴ Ă ӈ ݈ ೩ ี (premise)ણᇴĄ ௐ˟ᆸ ఢᆸĈซҖតᇴมሀቘទᏭఢ ̝ А ՙ ୧ І ੨ ၆ Ă ͽ ז Ч ఢ ̝ firing strength(ӈᝋࢦࣃ)ĂГӀϡ T-norm ࢷ᎕ྻზĂӈ ᏮࣃࠎٙѣᏮˢੈि̝ࢷ᎕Ĉ 2 , 1 ), ( ) ( × = = x y i wi µAi µBi ... (2) ௐˬᆸ ᝋࢦπӮĈѩᆸЧ༼ᕇࢍზྍఢ ࠹၆ٺٙѣఢ۞firing strength ּ̝ͧĄ 2 , 1 , 2 1 , 3 = = + i= w w w w O i i i ... (3) ௐαᆸ ඕኢଯኢᆸĈ 2 , 1 ), ( , 4 =w f =w px+q y+r i= O i i i i i i i ... (4) ̚{ pi,qi,ri}ࠎሀቘଯኢ̝ඕኢણᇴĂӈଯኢ ี(consequent)ણᇴĄ ௐ̣ᆸ ᏮᆸĈ݈ᆸੈिΐᓁͽࢍზᏮ តᇴࣃâтྋሀቘ̼̝ΑਕĈ 輸出值= ∑ ∑ ∑ = = i i i i i i i i w f w f w O5,1 ... (5) ANFIS ඕЪ˞݈㒝ёᙷৠགྷშྮ۞Ⴞ༛ё
ጯ௫ڱĂдˢቚቑּޢĂͧྵৌ၁Ꮾࣃᄃ ሀёଯҤࣃม۞ᄱमĂдՐᄱम۞̈π͞ ࿅̚Ăֹሀቘଯኢր̚۞ٙѣણᇴүዋ༊۞ አፋĄણᇴ۞࣒ϒ͞ڱߏଳϡ˟ล߱۞Ъёጯ ௫ႊზڱĈ(1)дੈཱིШ݈็۞ొ̶ĂЧᆸ༼ᕇᏮ ࣃـ݈็ҌௐαᆸޢĂᖣϤ̈π͞ଯҤڱ (Least squares estimate)ֽአፋଯኢีણᇴĄ(2)ᄱ मੈཱིుᆸਗ਼Ш็ጱҌௐ˘ᆸĂГӀϡੲࢫ ڱ(Gradient descent approach)Հາ݈೩ีણᇴĄᖣ Ϥ˟ล߱ጯ௫ԔĂANFIS ӈΞ࿅ᏮˢęᏮ ྤफ़ᄃˠᙷۢᙊ(̼ࠎሀቘ Ifęthen ఢԛё)ޙ ϲᏮˢęᏮ̝ߍडᙯܼĄ ͽώࡁտٙޙϲ̝ͪऱፆүրࠎּĂր ̝ᏮˢតᇴΒӣĈॡมT (ಏҜࠎџ)ăᅮͪณk k D (ಏҜࠎѺ༱ϲ̳͎͞)ă݈ഇˢ߹ณIk−1ă݈ ˟ഇᄊͪณSk−1,Sk−2ă݈ഇٸ߹ณOk−1ඈć˘ ࣎ᏮࣃĈٸ߹ณO ĄϤٺՏ࣎តᇴΒӣᇴ࣎k ఢٙᛳᔴᛳבᇴ̝ણᇴĂ༊ણᇴᄃఢᇴณ ͉кĂࡶۡତഴ͌ሀቘఢᇴٕᖎ̼ᔴᛳבᇴĂ ΞਕጱրপّᄃҖࠎೡ̙ԆፋĂ߇ืࡁᛉ ዋ༊۞ჸᙷ͞ڱĂᏮˢᄃᏮШณЪј̝Чඊ ྤफ̶़ᙷĂͽѣड़ഴ͌ણᇴ࣎ᇴٙౄј̝ᓄኑࢍ ზᄃྤफ़ᐼх۩ม̝Ąώࡁտଳϡሀቘഴڱ ჸᙷ̶ژ(Chiu, 1994)ՙؠሀё̚ሀቘఢऱ۞ ఢ࣎ᇴͽ̈́Чఢ̝ЪĂͽዋ༊гޙϲሀ ቘଯኢր̚۞ఢऱĄ
αăᑕϡ၁ּ
4.1 Ϯܝͪऱᖎ̬ ୶ͪګߏέ៉ௐˬ̂ګ̌ĂБܜ159 ̳֧Ă ߹ાࢬ᎕2,762 π̳֧͞ĂВѣˬ୧͚߹Ă̂႔ ࠎ୶ͪګ͚߹ܜ۰ĄϮܝͪऱҜٺ̂႔˯ ഫĂะͪડࢬ᎕763.4 π̳֧͞ĂӑͪҜᇾ 195 ̳͎Ă႕ͪҜᇾ245 ̳͎Ăѣड़टณ 2.357 ᆆ ϲ̳͎͞ĂгநҜཉтဦ4 ٙϯĄ Ϯܝࠎ˘кϫᇾͪऱĂͽă൴ă̳В ගͪࠎࢋϫᇾĂд߸ͪഇ֭ѣ֨߸ΑਕĂٺ˘ ਠॡഇϺฟٸֻϔிྼጵĄҋϔ઼ 53 ѐᎸޙԆ јޢĂӈॲፂͪऱྻϡ̣̂ૄώࣧ̈́ѐϡڱࠎ ૄĂࢎؠ˘ѐ̚Чॡഇ̝ͪऱͪҜࢨטѡ 25 0 25 50 75Km N ဦ4 ϮܝͪऱҜཉဦ 250 245 240 235 230 225 220 215 210 205 200 ဦ5 Ϯܝͪऱྻϡఢቢဦ ቢĂӈϮܝͪऱྻϡఢቢ(M-5 ఢቢ)Ăఢቢ̶ࠎ ˯ࢨă˭ࢨ̈́ᚑࢦ˭ࢨĂ˯ࢨ̝ؠཌྷࠎѣड़ᄊͪ ณٺᖳ࠳ېၗ̝ҲͪҜĂࢋᙯܼ֨߸ፆ үć˭ࢨߏѣड़ᄊͪณٺͪېၗ̝Ҳͪ ҜĂᙯܼ߸ͪ؞ޢ۞ϡͪᐼᄊćᚑࢦ˭ࢨߏ ѣड़ᄊͪณٺᚑࢦͪېၗ̝ҲͪҜĂᙯܼ ༊ॡ̝ගͪĄ ϤٺЧᇾ۞ϡͪᅮՐుѐᅍᆧĂࣧؠ̝M-5 ఢቢ̙̏னڶĂࠎჯͪऱֻ̝ͪड़ৈĂΔડ ͪྤԊٺϔ઼91 ѐࢦາᑭĂ M-5 ఢቢ 6 ͡˯ࢨϤ220 ̳͎೩ࠎ 235 ̳͎Ă࣒ϒ݈ޢ̝ ϮܝͪऱM-5 ఢቢтဦ 5 ٙϯĄ 4.2 ޙϲᐹᏴሀё ϤٺϮܝͪऱ።Ϋፆүྤफ़(1966-2001 ѐ) ̚Ăָ֭۞ٸ߹ณፆүྤफ़Ăࠎ˞ޢᜈംᇊ ݭͪऱፆүր̝ޙϲĂυืАՐפ˘நຐ۞(ָٕ۞)ٸ߹።ྤफ़үࠎϏֽሀё̝ቚྤ फ़ĄޙϲᐹᏴሀёॡĂࢵАᅮనؠϫᇾבᇴᄃࢨ טёĂГӀϡ็ႊზڱՐ႕֖ࢨט୧Іͷ྿ ͪᇴ̝̈ࢋՐ˭Ă࿅Νˬ˩ዶѐ̝ͪऱ ָٸ߹።ᄃ࠹ᙯ̝ͪऱटณត̼ྤफ़Ą̚ϫ ᇾבᇴᄃࢨטёࠎĈ (1)ϫᇾבᇴĈдᐹᏴ࿅̚Ăϡֽҿᕝՙඉ ͞९̝ᇾć၆ͪऱր҃֏Ăϫᇾבᇴݭёՙ ؠ˞ͪྤ۞Ӏϡ͞ёĄ˘ਠࢎؠͪऱፆүϫᇾ בᇴĂ࠰ഇ୕ਕ྿זᓁͪณ͌ăͪᇾ ̈Ă҃ͷࠎ˞ᔖҺд͌ᇴೀџ̰யϠ̂ณͪ҃ ౄјᚑࢦ۞ԿયᗟĂϺԓ୕ͪณਕ̶ͷ̙ ాᜈг̶ҶٺкџมĂͽഴ͌ͪ၆Чᇾ۞ϡͪ ̝ᑝĄώࡁտޙϲָ̼ϫᇾבᇴॡĂ҂ᇋϮ ܝͪऱѣஐͪᇄٙᅮ̝̳Вගͪᄃડٙᅮ̝ ྺຽϡͪีࢋϡͪᅮՐĂԓ୕ጐณπӮֻ ͪĂֹᚑࢦͪଐڶԼචĂ߇ଳϡͪᇴ۞ໄ هޙϲϫᇾבᇴݭёĂഇਕдѩϫᇾጱ˭ຩವ זָፆү͞ёĂͽѣड़ഴҲԿխຫεĄ ϫᇾבᇴĈ ) ) , 0 max( min( ) min( 2 36 1 i i i i i n D O D n ObjFunctio × − = ∑ = ̚ĂD ăi O ̶Ҿࠎௐ i џ̝ᅮͪณăٸ߹ณĂi i n ࠎௐ i џ۞᎕ͪџᇴĂё̚ͽπֽ֝͞ి ᆧ̂בᇴࣃĂΞᕖ̂ͪณड़ᑕĂᖣѩᔖҺͪ ࿅ٺะ̚ćѩγĂn Ξᕖ̂ాᜈͪड़ᑕĂᔖҺi ాᜈͪଐԛ൴ϠĄѩγĂϤٺᄐ൴ͪณޝ̈Ă ߇ώࡁտ̚ᄐ൴ณنர̙ࢍĄ (2)ր̝ࢨטёΒ߁Ĉ ాᜈ͞ёĈր̚Чͪ̍ඕၹۏӮื Ъͪ߹۞πᏊ୧ІĂӈˢ߹ณඈٺ ߹ณĄ҃ͪऱإื҂ᇋᄊͪड़ᑕĂӈ ߹ณඈٺˢ߹ณഴΝᄊͪณ̝ត̼Ą ᄊͪࢨטёĈٺፆүഇมͪऱटณื̬ ٺѣड़टณ̰Ăӈӑटณŷͪऱटณŷ ̂ऱटĄϮܝͪऱ̂ऱट˜ߏॲፂ ͪӀཌშ৭ྤफ़(ၟҌϔ઼ 91 ѐ 4 ͡ ͤ)ĂϤࣧАࢍ൪ѣड़टณ 251.88 Ѻ༱ ϲ̳͎͞ࢫҌϫ݈ѣड़टณ235.745 Ѻ ༱ϲ̳͎͞Ą ͪऱ۞ࢋΑਕдٺჯͪऱܜഇග ͪᄃ֨߸̝ᘦؠّĂЯѩдፆүඉர ˯Ă̙آனߜͪѐϡͪณٕᖳͪѐ ள૱ᐼͪඈଐԛĂ߇ࢨטՏ36 џͪऱ ፆүޢĂͪऱटณ̙ਕᄃፆүܐഇ̝ऱ ट࠹म͉̂Ąώࡁտऱटត̼ࢨטࢎ ࠎ10%Ą 限制條件： 0 36 0 0 1 1 . 1 9 . 0 745 . 235 0 0 . 50 S S S S S O I S S i i i i i ≤ ≤ ≤ ≤ = − + = − ̚ĂS ăi I ̶Ҿࠎௐ i џ۞ͪऱटณăˢ߹ณĂi 0 S ࠎܐؕᄊͪณĄϤٺͪऱఢቢٺܧѴഇഇ มĂ˯ࢨă˭ࢨ̈́ᚑࢦ˭ࢨͪҜࣃ࠹मྵ̂Ăࡶ ͽௐ1 џүࠎؕፆүџĂऻЯតજቑಛྵ̂ ֹ҃ፆүඕڍஎצܐؕͪऱटณᇆᜩć߇Լͽௐ 19 џ(7 ͡ௐ 1 џ)үࠎؕፆүџĂྍॡഇࠎ֨ ߸ഇĂͪऱͪҜྵҲĂтѩܐؕᄊͪณྵ̙ᇆᜩ ޢᜈፆүඕڍĄώࡁտણ҂Ϯܝͪऱ࿅ΝͪҜᄃ M-5 ఢቢ˭ࢨࣃĂؕፆүџ̝ᄊͪณࢎࠎ 50 Ѻ༱ϲ̳͎͞Ą 4.3 ͪऱፆүఢቢᄃሀቘఢऱ̝ᖼೱ έ៉гડனѣͪऱдፆү˯кଳྻᖼఢቢ ͞ёĄ็̝ఢቢፆү͞ё˜ߏϏֽˢ߹ณෛ ࠎனѣͪऱᄊͪณ̝בᇴĂΐ˯࿅Ν߹ณᐂ̝ ᔌ๕ĂГֶᅮͪณкဿĂᑢࢎ˘ΞᔖҺϏ ֽॡഇ̂ณͪĂ˫Ξϫ݈Яഴֻ͔͌ͪٙ ̝̙ӀᇆᜩഴҌ̈۞ፆү͞ёĄϤٺఢቢፆү ߏଳ“πӮ”ྵр۞͞ёซҖͪऱፆүĂੵݒߙֱ ͪ͛ณតளྵ̂۞ଐڶ(ּтᜈԿ)γĂఢቢ ࠰Ξჯͪऱ۞ᘦؠፆүĂͷఢቢࢎؠ̝ܐĂ ણ҂˞ధкछጯ۰۞ۢᙊᄃགྷរĂࡶਕ็ ఢቢፆү̝छۢᙊᄃ࿅Νࡁտٙޙϲ̝ം ᇊݭͪऱፆүրඕЪĂΞৼˢొ̶ఢቢٸ ̝ͪᐹᕇĂֹሀё่̙Βӣ࿅Ν።Ϋྤफ़ٙ ᔳӣ۞ྤੈĂТॡ˵౯˞छۢᙊĂΞՀ ംᇊгซҖፆүĂ֭೩ֻͪऱგந۰ϒቁă
Ξያ۞ᏲೈֶፂĄ ЯѩĂώࡁտ̝ࢦᕇӈߏࡁᛉͪऱፆүఢቢ ᄃሀቘఢۢᙊऱ̝ม۞ᖼೱ͞ёᄃ፟טĂఢ ቢܑ̝ٙᄊٸᇾᖼೱࠎఢĂޙϲ˘ሀቘఢ ۢᙊऱĄֶፂϮܝͪऱM-5 ፆүఢቢ̝ఢؠĂ ͪऱࢋߏֶፂ༊ॡ۞ͪऱͪҜᄃ˭ഫᅮͪณ ซҖፆүĈ ͪऱͪҜᇾ˯ࢨॡĂܑϯͪऱࠎ ᖳͪېၗĂᑕෛ၁ᅫᅮࢋႽณ൴Ą ͪऱͪҜᇾд˯ă˭ࢨ̝มॡĂܑϯ ͪऱхͪϒ૱ĂЧᇾ۞ϡֶͪࢍ൪੨ͪ ณֻͪĄ ͪऱͪҜᇾд˭ࢨᄃᚑࢦ˭ࢨ̝ม ॡĂܑϯͪऱѣᅅߜԿன෪ĂੵЧᇾ ۞ϡֶͪࢍ൪੨ͪณֻ̙̟ͪᆧΐĂࠎ ЯᑕΞਕ̝ᜈߜԿĂͪऱგநಏҜᑕ Аםથ੨ͪณഴֻନ߉Ą ͪऱͪҜᇾҲٺᚑࢦ˭ࢨॡĂܑϯͪ ऱѣᚑࢦͪଐԛĂϡͪࢍထ੨ ͪณ˛ј੨ٸĄ ᓝּֽᄲĂ༊ͪऱͪҜᇾҲٺᚑࢦ˭ࢨ ॡĂఢΞࢎࠎĈ༊ॡมࠎௐA ࣎ॡഇăͪऱͪ1 Ҝࣃ(L)ࡶҲٺྍॡഇፆүఢቢ̝ᚑࢦ˭ࢨࣃ (B )Ăٸ߹ณ( R )ࠎᅮͪณ( D )۞Ѻ̶̝˛1 ˩ĄТநĂՏ࣎ॡഇ࠰Ξॲፂѩٸͪఢؠޙϲྍ ॡഇٙ၆ᑕ̝ఢт˭Ĉ ) 7 . 0 ( ), ( : ) 7 . 0 ( ), ( : 2 ) 7 . 0 ( ), ( : 1 2 2 2 1 1 1 k k kandL B Then R D A T If k Rule D R Then B L and A T If Rule D R Then B L and A T If Rule = ≤ = = ≤ = = ≤ = M ఢቢፆүд၁ᅫᑕϡ˯ᔵѣֹϡ͞ܮă៍ هᖎಏඈᐹᕇćҭдᒉྻፆү࿅̚Ăͪऱˢ߹ ณ̈́Чᇾ۞ᅮՐณצˠࠎ̈́ҋЯ৵ඈᇆᜩ࠰ ࠎតજณĂͷఢቢቑಛྵ̂Ăдֹϡ˯ྵͻᇅ ّĂڱซҖྵჟቁ̝ፆүĂЯѩ၁ᅫፆүॡੵ ֶፂ࠹ᙯᐂγĂυᅮГᖣϤፆүˠࣶܜഇ۞གྷ រ᎕ᄃംᇊҿᕝĂ͞ΞჯͪऱྻᖼᄃቁܲЧ ᅮՐ̝םአĄপҾߏள૱ͪ͛ېڶ˭Ăѣ੨Ъ ˠࠎۢᙊҿᕝซҖአ༼ٸͪĂ̖Ξࣘᜪͪऱͪྤ ᇅّአޘΑਕᄃͪऱщБّĄ҃छۢᙊᄃ ᖳಱགྷរ̙ٽᒔĂ˵̙ٽᖼೱࠎఢݭёĂ Яѩଳϡሀቘநኢֽநˠᙷ۞ۢᙊᄃទᏭଯ ኢ࿅ٙ̚Βӣ۞ሀቘّĂᐹᕇӈࠎତܕˠᙷ ۞ޥ҂ҖࠎĂྵटٽඕЪछۢᙊĄ ሀቘநኢߏϤ Zadeh(1965)ٙ೩Ă၆ٺ к̮ኑᗔ̝ሀቘன෪Ăග̟ྵࠎᘦઉ̝ೡĄд ந၁ᅫયᗟॡĂࢋߏ఼ะЪĶܧѩӈكķ ̝ ၆ ᔴ ᛳ ᙯ ܼ ΐ ͽ ᕖ · Ă Ӏ ϡ ᔴ ᛳ ב ᇴ (Membership Function)۞៍هĂؠณג൪̙ቁؠ ّયᗟ̝ሀቘّኳĂЯѩ၆ٺୃ̙ٕېڶሀ ቘ̝યᗟĂ೩ֻ˞˘࣎ྵЪநΞҖ۞ྋՙ͞ёć дࡊጯᄃጯఙ۞ࡁտ˯ĂֹϡሀቘநኢΞந ᄬຍ̶ٕژ۞ೡّᄬ֏Ăྋՙ็ะЪٕநኢ ٙڱೡ۞ன෪ᄃયᗟĄܕѐֽĂ̏జјΑᑕ ϡٺ̙Т۞ͪྤયᗟ(Russell and Campbell, 1996ćShrestha et al., 1996ć Dou et al., 1999ć Dubrovin et al., 2002)Ą ˘ਠ҃֏ĂሀቘఢΞϤछ೩ֻăۢᙊᕜ פٕགྷϤྤफ़প̶ّᙷயϠĄдώࡁտ̚ߏϮ ܝͪऱM-5 ፆүఢቢ̝ٸͪఢؠ̼ࠎఢĂࢵА ࢎؠሀቘఢ۞Ꮾˢតᇴࠎॡม̈́ͪऱͪҜĂᏮ តᇴߏ̙ТᅮՐ˭۞ٸͪณĂࢎؠఢݭё т˭Ĉ ) ( ), A is a A is a A is (a
1 i,1 2 i,2 k i,k Then Ri
If • •L• ̚a ࠎௐ k ࣎ᏮˢតᇴĂk Ai,kߏௐk ࣎តᇴ ۞݈೩ีĂ֭ͽᔴᛳבᇴMi,k ۞ሀቘݭёܑனĂ ҃R ߏௐ i ࣎ఢ۞ଯኢีĂϺߏͽሀቘݭёi ܑனĄдᏴፄሀቘఢֹϡ۞ᔴᛳבᇴॡĂᅮ҂ ᇋዋ༊۞ݭёĂ૱ϡ۞ѣˬ֎ԛבᇴăୗԛב ᇴăᛗݭٕבᇴඈ(тဦ 6 ٙϯ)Ąώࡁտࠎ ੨Ъፆүఢቢ̚ĂͪऱͪҜҲٺᚑࢦ˭ࢨॡֶࢍ ထ੨ͪณ˛ј੨ٸ۞ٸͪఢؠĂᏴϡS ݭᔴᛳב ᇴ۞ሀቘݭёĂГ࿅ણᇴ̝አፋĂֹЪ၁ ᅫٸͪଐԛĂтဦ7 ٙϯࠎᏮˢតᇴ─ͪҜ̝ᔴ ᛳבᇴݭёĂనௐi џ M-5 ፆүఢቢ̝ᚑࢦ˭ ࢨࣃࠎ 60Ă༊ͪऱͪҜҲٺ 60Ăྍॡഇ۞ٸ ֶͪࢍထ੨ͪณ˛ј(0.7)੨ٸĄЯѩЧџ࠰Ξ
1 0 µ(x) 1 0 µ(x) 1 0 µ(x) 1 0 µ(x) a c c b x x (a) a a c d b x c x (b) (c) (d) σ slope = −b/2a ဦ6 Чԛё۞ᔴᛳבᇴ 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 10 20 30 40 50 60 70 80 90 100 S ဦ7 Sݭᔴᛳבᇴݭё ࢎؠЧҋ۞ఢĂఢؠྍџͪऱͪҜࡶҲٺྍ џM-5 ፆүఢቢ۞ᚑࢦ˭ࢨࣃĂྍџٸͪณֶ ࢍထ੨ͪณ˛ј(0.7)੨ٸĄ 4.4 ޙϲ ANFIS ሀё ࠎֹሀёޙၹॡણᇴΞ྿זָ̼ېၗĂд ።Ϋྤफ़ᇴณ·̶۞୧І˭Ă఼૱ٙѣΞᇴ ፂડ̶ࠎቚăរᙋᄃീྏˬ࣎࠹̢ϲ۞ྤफ़ ĄࢵА็ႊზڱٙଯՐָ̝ٸ߹።(В 36 ѐ 1,296 ඊྤफ़)̶јˬొ̶Ă̚ 828 ඊྤफ़ ࠎቚቑּĂүࠎANFIS ሀёአፋણᇴ̝ቚྤ फ़ĂΩγ216 ඊྤफ़ϡͽរᙋሀёߏӎዋϡٺ၁ ᅫͪऱፆүඉர̝ՙؠĂޢ252 ඊྤफ़ोֽ ซ Җ ଯ Ҥ ീ ྏ ĄA N F I S ሀ ё Ξ ॲ ፂ ॡ ม 5 10 15 20 25 30 35 30 35 40 45 50 55 60 65 RMSE ဦ8 RMSE ࣃᄃఢᇴ̝ᙯܼ k T (ಏҜࠎџ)ăᅮͪณD (ಏҜࠎѺ༱ϲ̳͞k ͎)ă݈ഇˢ߹ณIk−1ă݈˟ഇᄊͪณSk−1,Sk−2ă ݈ഇٸ߹ณOk−1ඈᏮˢតᇴĂଯҤٸ߹ณO Ą k ANFIS ሀё̚ଳϡˬሀቘఢऱซҖͧ ྵĂ̶ҾࠎĈ(1)ͽ GA ٙᐹᏴָ̝ٸ߹።ү ࠎቚᇹώᄃᇾ۞Ăଳϡሀቘഴڱჸᙷ၆።Ϋྤ फ़ ޙ ϲ Ꮾ ˢ ត ᇴ ᄃ Ꮾ ត ᇴ ม ۞ ሀ ቘ ఢ ऱ (GA)ć(2)Ϯܝͪऱ M-5 ፆүఢቢᖼೱࠎሀቘ ఢۢᙊऱ(FRB)ć(3)ͽ˯˟ఢऱ̝ඕЪ (GA & FRB)Ąͽ˭̶Ҿಶˬሀቘఢऱ̝ඕڍ ซҖͧྵĄ (1)ሀቘఢऱ(GA)ĈώఢऱӀϡჸᙷ͞ڱ Տ˘ඊϤᏮˢШณᄃᏮШณЪј۞ྤफ़ΐ ͽ̶ᙷ֭ྻϡٺሀቘIf-then ఢ̚Ă҃ͽዋ༊ гޙϲሀቘଯኢր̚۞ఢऱĄдՙؠఢᇴ ϫॡĂֶፂሀёଯҤࣃᄃৌ၁ࣃ̝ᄱमࣃֽՙ ؠĂтဦ8 ٙϯĂఢᇴࠎ 17 ॡĂRMSE ࣃ ̈Ă߇ώఢऱᏴϡ17 ࣎ሀቘఢĄ (2)ሀቘఢۢᙊऱ(FRB)ĈϮܝͪऱ M-5 ፆүఢቢᖼೱࠎఢݭёĂߛၹANFIS ሀё̝ ሀቘఢۢᙊऱĂГӀϡშྮ̝ҋԧአዋĂు႙ አፋዋ༊̝ણᇴĂͽЪͼሀቘଯኢր̚Ꮾˢę ᏮมᙯܼĄՏ˘џ۞ٸͪఢؠ࠰ࢎؠ˘୧ሀቘ ఢ(т 4.3 ༼ٙϯ)Ă߇ FRB ̚Βӣ 36 ୧ሀቘఢ Ą (3)ඕЪ˟ఢऱ(GA & FRB)ĈӈඕЪ˞ GA ሀቘఢऱᄃ FRB ሀቘఢۢᙊऱĂВࢍѣ 53(17+36)୧ሀቘఢĄ
̣ăඕڍᄃኢ
5.1 ඕڍͧྵણᇴ
GSI ͪᇴ
ѣड़۞ͪᇴυืਕቁ̷гͅᑕͪ ᐛதăૻޘͽ̈́ՏѨͪ۞ؼॡĄώࡁտଳϡ GSI(Generalized Shortage Index)(HsuĂ1995)үࠎ ޢᜈ̝ፆүඕڍͧྵֶፂĂЯࠎGSI ၆ԿְІ ̝ͪณᄃాᜈड़ᑕྵࠎୂຏĂЯѩ၆ٺԿְ І̝ෞҤྵࠎމ៍ĂͷՀΞૻአ̝ͪۤົј ώĄGSI ̝ؠཌྷт˭Ĉ k N i i i DY DPD N GSI NDC DDR DPD ∑ ∑ = × = × = 1 0 0 ) 100 ( 100 ) ) ( (
̚ĂDPD (Deficit Percent Day Index)ܑௐ ii
ѐ̝DPD ࣃĂΒ߁ͪૻޘᄃాᜈّćDDR ࠎ џͪதĂӈ(ྍџ۞ϫᇾᅮͪณůྍџٸ߹ณ)/ ϫᇾᅮͪณćNDC ࠎྍְͪІϫ݈̏᎕̝ా ᜈͪџᇴćDY ߏௐ i ѐ̝џᇴ(36)ćN ࠎͧྵi ̝ѐᇴćK ࠎણᇴĂ఼૱నࠎ 2Ą RMSE ࣃ ੵ˞ͧྵͪऱፆү၆ͪଐԛ̝ԼซĂࠎͧ ྵ ANFIS ሀё̝ቁّĂΩᏴϡӮ͞ॲᄱमࣃ (Root Mean Square Error, RMSE)үࠎ ANFIS ീ ඕڍᄃቚᇾݭ(GA ᐹᏴඕڍ)̝ͧྵᇾĄ RMSE ࣃດٕ̈ດᔌܕٺ 0 ܑϯሀёດቁĄ ؠཌྷт˭Ĉ
(
)
0.5 1 2 ˆ − = ∑ = N i i i N Q Q RMSE ̚Qˆ ăi Q ̶Ҿࠎௐ i ഇଯҤٸ߹ณăৌ၁ٸ߹i ณ(дѩࠎ GA ᐹᏴ̝ٸ߹ྤफ़)ĂN ࠎྤफ़ᕇ ᇴĄ 5.2 ඕڍ 5.2.1 ็ႊზڱᐹᏴඕڍ ώࡁտଳϡ̝Ꮾˢྤफ़Β߁Ĉ2001 ѐࢍ൪ᅮ ͪณ۞1.2 ࢺྤफ़(ᓁᅮͪณࠎ 1329.385 Ѻ༱ϲ͞ ̳͎)Ă̈́Ϯܝͪऱҋ 1966 Ҍ 2001 ѐВ 36 ѐ̝ ።Ϋџ߹ณྤफ़Ą ࢵАӀϡ።Ϋྤफ़ሀᑢϮܝͪऱͽ M-5 ఢ ቢซҖͪऱፆүĂГͽ็ႊზڱຩವָٸ߹ ።Ăͧྵ˟͞ڱ̝ፆүඕڍ(ဦ 9)ĄϤ GSI ͧྵဦΞځព࠻Ĉˢ߹ณྵ̂۞ᖳͪѐॡĂ˟ ͞ڱ۞ͪଐԛ࠹̙ᅈĂዶ۞ඕڍពϯ GA ႊზඕڍ۞ͪᇴ࠰̈ٺ M-5 ఢቢፆүඕ ڍĂӈಶGSI ͪᇴ҃֏ĂGA ᐹᏴඕڍځព ᐹٺM-5 ఢቢፆүĂӈдώࡁտٙనؠ̝ϫᇾב ᇴጱ˭ĂGA Ξຩವזͪณăͪџᇴ̶ ĂͽᔖҺໂბָ̝ͪፆүඉரĄ߇GA ٙ ᒔ̝நຐٸ߹ณ።Ξᑕϡٺޢᜈംᇊݭ ͪऱፆүր̝ቚྤफ़Ą 5.2.2 ANFIS ሀёീඕڍ ܑ1 Еቚăរᙋăീྏˬล߱˭Ă̙Т ፆү͞ё۞ඕڍĂ̶ҾࠎM5(ͽ M-5 ఢቢซҖͪ ऱ ፆ ү)ćGA(ͽ็ႊზڱຩವָٸ߹። )ćANFIS(ͽ GA ᐹᏴ̝ٸ߹።үࠎቚྤ फ़ĂГӀϡˬ࣎ሀቘఢऱซҖͪऱፆү)Ąඕڍ Ξ࠻ĈಶGSI ᇴ҃֏ĂGA ຩವඕڍځពᐹ ٺM-5 ఢቢĂЯѩ GA ᐹᏴָ̝ٸ߹። үࠎANFIS ሀё۞ቚᇾݭĂ ANFIS ሀё ኢߏࣹఢऱдቚăរᙋăീྏˬลܑ̝߱ ன࠰ྵ็M-5 ఢቢፆү۞ඕڍྵָĂӈ ANFIS ሀё۞ቁΞд႕֖˭ഫᅮͪ۞݈೩˭ĂଯՐͪऱ ָٸ߹ณĂ֭ਕѣड़ᔖҺͪะ̚۞ᚑࢦԿ ଐԛĄ ѩγඕڍϺពϯ ANFIS ሀё۞ˬఢऱ ̚ĂඕЪ።ΫְІ۞གྷរᄃछፆүۢᙊ̝ఢ ऱ(GA & FRB)Ăྵ˟ఢऱ҃֏Ă۞ቁѣ ड़гԼචፆүඕڍĂӈѩఢऱٙޙϲ̝ANFIS ሀёംᇊгซҖͪऱፆүĄ ੫၆ሀёീྏล߱(1995-2001 ѐ)ซҖෞͧ (тܑ 2 ٙϯ)Ă̚ 1996 ѐѣԿன෪Ăྍѐፆ үඕڍ̚ANFIS ሀё۞ˬ࣎ఢऱĂ GSI ࣃ ࠹ྵٺM-5 ఢቢፆү࠰ѣځពซՎĂᔵᓁͪ ณྵM-5 ఢቢፆүࠎ̂ĂࣧЯࠎώࡁտٙ೩̝ ፆү͞ёϫ۞ߏԼචͪଐԛะ̝̚ᚑࢦԿ ன෪Ă߇ԓ୕Կॡ۞ͪณ̶ҌྵкџĂ Я ѩ ᓁ ͪ ณ ࠹ ၆ ྵ к Ą ѩ γ Ă ࠎ ซ Җ ͪ ऱ6000 5000 4000 3000 2000 1000 0 2500 2000 1500 1000 500 0 GSI GA M-5 53 56 59 62 65 68 71 74 77 80 83 86 89 ဦ9 GA ፆүᄃ M-5 ఢቢፆү̝ GSI ͧྵ ܑ1 ˬล߱˭ M-5ăGAăᄃˬ ANFIS ሀёඕڍͧྵ ANFIS M-5 GA ሀቘఢऱ (GA) ሀቘఢۢᙊऱ (FRB) ඕЪ˟ఢऱ (GA & FRB) ѐ GSI GSI RMSE GSI RMSE GSI RMSE GSI ቚล߱ 1966-1988 394 33 32 157 40 79 39 135 រᙋล߱ 1989-1994 982 54 27 238 35 298 36 185 ീྏล߱ 1995-2001 633 67 31 171 33 131 29 47 πӮࣃ 670 51 30 189 36 169 35 122 ܑ2 ീྏล߱ M-5 ఢቢፆүᄃˬఢऱ̝ ANFIS ሀёඕڍͧྵ ANFIS ሀቘఢऱ (GA) ሀቘఢۢᙊऱ (FRB) ඕЪ˟ఢऱ (GA & FRB) M-5 ఢቢፆү ѐ (1) (2) (3) (1) (2) (3) (1) (2) (3) (1) (2) (3) 1995 249 21 32 310 19 40 263 21 27 42 3 1 1996 611 31 417 648 32 733 460 26 143 602 32 4256 1997 359 21 249 248 18 16 272 22 37 218 12 66 1998 258 17 32 187 16 6 229 19 22 35 8 3 1999 457 21 379 372 21 44 415 24 82 290 14 49 2000 330 24 45 294 23 38 244 18 9 119 13 56 2001 292 23 46 259 19 39 164 16 6 11 1 0 ᓁ πӮࣃ πӮࣃ ᓁ πӮࣃ πӮࣃ ᓁ πӮࣃ πӮࣃ ᓁ πӮࣃ πӮࣃ 2556 23 171 2318 21 130 2050 20 47 1318 12 633 (1)ᓁͪณ(Ѻ༱ϲ̳͎͞)ć(2)ͪџᇴć(3)GSI ٸ߹ณീྏࣃᄃ၁ᅫࣃ(GA ᐹᏴָ̝ٸ߹። )̝ͧྵĂҋរᙋăീྏล߱ЧᏴ̣ѐྤफ़ᘱ јᔌ๕ဦĂဦ10ăဦ 11 ӈࠎඕЪ˟ఢऱ(GA & FRB)̝ ANFIS ሀёซҖͪऱٸ߹ณଯҤ̝រ ᙋᄃീྏล߱ᔌ๕ဦĂΞͽ࠻ANFIS ሀё၆ٺ ٸ ߹ ณ ۞ ീ ѣ ̙ ̝ ј ڍ Ă ̙ ኢ д ᔌ
20 40 60 80 100 120 140 160 180 0 100 200 300 400 500 600 verification ( ) 7 ( ) GA ANFIS ဦ10 ඕЪ˟ఢऱ̝ ANFIS ሀёଯҤͪऱٸ߹ณ̝រᙋล߱ᔌ๕ဦ 450 400 350 300 250 200 150 100 50 0 20 40 60 80 100 120 140 160 180 ( ) ( ) testing GA ANFIS ဦ11 ඕЪ˟ఢऱ̝ ANFIS ሀёଯҤͪऱٸ߹ณ̝ീྏล߱ᔌ๕ဦ ๕ٕߏቁத͞ࢬӮܑன̙Ą ፋវ҃֏ĂඕЪ˞࿅Ν።Ϋྤफ़ٙᔳӣ۞ͪ ͛ྤੈͽ̈́Ϥछགྷរᄃፆүۢᙊٙࢎؠ۞ఢ ቢ̝ఢऱĂ·̶г൴೭˞ͪऱፆүր̝ം ᇊĂANFIS ሀёځពԼච M-5 ఢቢፆүͪะ̚ ̝ଐԛĂᙋځANFIS ሀёᖣϤംᇊݭଠט̝፟ט ଠטͪऱͪҜᄃٸ߹ณĂΞ೩ֻͪऱგநԊՙ ؠϏֽፆү̝ણ҂ඉரĄ
ᔁ
ᄫ
ώࡁտᄋҖ߆ੰ઼ࡊົྃӄొЊགྷĂࢍ൪ በཱིNSC90-2313-B-002-323ćࡁտഇมٚᄋགྷᑻ ొͪӀཌΔડͪྤԊՂᜠϔԊܜăϮܝͪऱგ ந͕̚ᖎߌཏЇ೩ֻᚗෳޙᛉ̈́к͞םӄĂᖰ ѩ׀ᔁԢĄણ҂͛ᚥ
1. Ѧုᅞăเ͛߆ăเׂĂ“ሀᑢڱдͪऱፆ үఢቢ˯̝ᑕϡ”Ăέ៉ͪӀĂ48(4):53-63, 2000Ą 2. ˜ᯂĂᑕϡԔதજၗఢထטࢎૻטّͪऱ ፆүఢቢ̝ࡁտĂ̚Ꮈ̍ࡊԫࡁտ൴णૄ ܛົĂ1997Ą 3. ܘߒĂሀቘఢထநኢᄃᐹᏴڱٺͪऱፆү ̝ࡁտĂ઼ϲέ៉̂ጯϠۏᒖဩր̍ࡁ տٙჇ̀ኢ͛Ă2003Ą 4. षѐăเୂംĂņሀᑢੜͫڱՙؠ͟͡ሔ ͪऱָఢቢ̝ᑕϡŇĂ˝˩ѐޘྺຽ̍ ࡁົኢ͛ะĂpp.907-916Ă2001Ą 5. ૺౢăเཌྷăୖॢᅛĂņሀቘଯኢሀё ̝ޙϲ̈́ϡٺͪ͛ր̝ࡁտŇĂ઼̚ྺ ຽ̍ጯಡĂ39(1):71-83Ă1993Ą 6. ૺᚊࡌăૺౢĂ“ംᇊݭͪऱӈॡፆүଠט ր”Ă઼̚ྺຽ̍ጯಡĂ45(4):18-30Ă 1999Ą 7. ୖॢᅛăૺౢăౘޜራĂņኑЪႊზᙷৠ གྷ-ሀቘଯኢሀёᑕϡٺ߸ͪീŇĂ̚රͪ ˿ܲጯಡĂ31(3)Ă2000Ą 8. ధ͌༉Ăͪऱፆүఢቢયᗟ۞ሀёᄃྋڱĂ ઼ϲέ៉̂ጯྺຽ̍ࡁտٙჇ̀ኢ͛Ă 2001Ą 9. ోॎăૺڠĂņᑕϡԔதજၗఢထٺϮ ܝͪऱྻᖼ̝ࡁտŇĂௐ˟بͪӀ̍ࡁ ົኢ͛ะĂpp.125-141Ă1984Ą 10. ౘᙶтĂ̂ϥ˭ഫአᄊͪѰटณనࢍᄃࢲ ᐍ̶ژ̝ࡁտĂ઼ϲέ៉̂ጯྺຽ̍ࡁտ ٙჇ̀ኢ͛Ă2001Ą 11. ౢĂ็ႊზڱ̝ࡁտ̈́ٺ̼ͪ͛ጯ ሀё̝ᑕϡĂ઼ϲέ៉̂ጯྺຽ̍ࡁտٙ Ⴧ̀ኢ͛Ă1994Ą 12. เॎཐĂഅ͛ͪऱրָᒉྻᄃࢲᐍ̶ ژĂ઼ϲέ៉̂ጯ˿̍͢ࡁտٙჇ̀ኢ ͛Ă1995Ą 13. Ꮒܛੑăૺ։ϒĂņкϫᇾͪऱָፆүሀ ё̝ޙϲᄃᑕϡŇĂᄂ៉ͪӀĂ46(1)Ă1998Ą 14. Becerikli, Y., A. F. Konar, T. Samad,“Intelligent optimal control with dynamic neural networks”, Neural Networks, 16(2): 251-259, 2003.
15. Cai, X. M., D. C. Mckinney, and L. S. Lasdon, ņSolving nonlinear water management models using a combined genetic algorithm and linear programming approach “Advances in Water Resources”, 24(6): 667-676, 2001.
16. Chandramouli, V., K. A. Kuppusamy, and K. Manikandan, “Study on water sharing in a multi-reservoir system using a dynamic programming-neural network model”, Interna- tional Journal of Water Resources Development, 18(3): 425-438, 2002.
17. Chang, F. J., S. C. Hui, and Y. C. Chen, “Reservoir operation using gray fuzzy stochastic dynamic programming”, Hydrological Processes, 16(12): 2395-2408, 2002.
18. Chang, L. C. and F. J. Chang, “Intelligent control of modeling of real time reservoir operation”, Hydrological Processes, 15(9): 1621-1634, 2001.
19. Chang, F. J., L. Chen, and L. C. Chang, “Optimizing the reservoir operation rule curves by genetic algorithms”, Hydrological Processes (accepted).
20. Chang, Y. T., L. C. Chang, and F. J. Chang, “Intelligent control for modeling of real time reservoir operation: Part II ANN with operating curves”, Hydrological Processes (in press). 21. Chiu, S. L., “Fuzzy model identification based
Fuzzy Systems, 2(3): 267-278, 1994.
22. Davies, R. M., and J. Watton, “Intelligent control of an electrohydraulic motor drive system”, Mechatronics 5(5): 527-540, 1995. 23. Dou C, W. Woldt, and I. Bogardi, “Fuzzy
rule-based approach to describe solute transport in the unsaturated zone”, Journal of Hydrology, 220(1-2): 74-85, 1999.
24. Dubrovin, T., A. Jolma, and E. Turunen, “Fuzzy model for real-time reservoir operation”, Journal of Water Resources Planning and Management, 128(1): 66-73, 2002.
25. Holland, J. H., “Adaptation in natural and artificial systems”, 2nd ed., Mass. Inst. of Technol., Cambridge, 1975.
26. Hsu, S. K., “Shortage indices for water- resources planning in Taiwan”, J. of Water Resources Planning and Management, 121(2): 119-131, 1995.
27. Jang, J-S R., “ANFIS: Adaptive network based fuzzy inference system”, IEEE Trans. On Syst. Man and Cyber., 23(3): 665-685, 1993.
28. Kelman, J., J. R. Stedinger, L.A. Cooper, E. Hsu, and S.Yuan, “Sampling stochastic dynamic programming applied to reservoir operation”, Water Resources Research, 26(3): 447-454, 1990.
29. Lin, C. L., and H. W. Su, “Intelligent control theory in guidance and control system design: an overview”, Proceedings of the National Science Council, ROC 24(1): 15-30, 2000. 30. Mantawy, A. H., Y. L. Abdel-Magid, S. Z.
Selim, “A new genetic-based tabu search algorithm for unit commitment problem”, Electric Power Systems Research, 49(2): 71-78, 1999.
31. Perera B. J. C., and G. P. Codner, “Computational improvement for stochastic dynamic programming models of urban water supply reservoirs”, J. American Water Resour. Assoc., 34(2): 267-278, 1998.
32. Russell, S. O., and P. F. Campbell, “Reservoir operating rules with fuzzy programming”, J. of Water Resources Planning and Management, 122(3): 165-170, 1996.
33. Shrestha, B. P., L. Duckstein, and E. Z. Stakhiv, “Fuzzy rule-based modeling of reservoir operation”, J. of Water Resources Planning and Management, 122(4): 262-269, 1996.
34. Thomas, L.C., T.W. Archibald, and K.I.M. McKinnon, “An aggregate stochastic dynamic programming model of multiple reservoir systems”, Water Resources Research, 33: 333-340, 1997.
35. Wang, Q. J., “The genetic algorithm and its application to calibrating conceptual rainfall- runoff models”, Water Resources Research, 27(9): 2467-2471, 1991.
36. Zadeh, L. A., “Information and control”, Fuzzy Sets, 8, 1965.
ќቇ͟ഇĈϔ઼93 ѐ 9 ͡ 20 ͟
࣒ϒ͟ഇĈϔ઼93 ѐ 12 ͡ 21 ͟
