ISIHLUNGI SESIHLANGO SAMABUSA EDIZISI

Ngokuthambekela kwemishini yemishini ekuqondeni okukhulu, okukhaliphile nokunemba okuphezulu, ukusetshenziswa kwezingxenye ezijikelezayo, njenge-roller bear, kuye kwathuthukiswa kakhulu ukuze kuzuzwe ukudluliswa kwamandla, ukulungiswa kwendawo kanye nezinye izinjongo. Uma zonakalisiwe noma zehluleka, ukuphepha kokusebenza kwemishini eyimishini nenzuzo yokukhiqiza izothinteka. Kodwa-ke, ngenxa yokuma okukhethekile kwalezi zingxenye ezijikelezayo, kuba nzima kakhulu ukucwaninga nokwahlulela isimo sezempilo semishini, futhi izindlela zangaphambili ezithembele kubantu noma ulwazi ngeke zisasebenza. Ngakho-ke, ukuthuthukiswa kwendlela yokuthola okuhlakaniphile kanye nendlela yokuxilongwa ukuze kusetshenziswe ukuqapha kwezempilo kwemishini sekuyisihloko socwaningo esishisayo.

Ngokuthuthuka okusheshayo kobuhlakani bokwenziwa, izindlela zokufunda zomshini ziya ngokuya zenza ukuxilonga okuhlakaniphile kwemishini kufezeke futhi kuphumelele, njengokufunda kokuqinisa (RL) [1], [2], amanethiwekhi adversarial akhiqizayo (GAN) [3], i-autoencoder. (AE) [4] kanye nomshini we-vector wokusekela (SVM) [5], [6], [47]. Phakathi kwazo, i-SVM iyi-algorithm yokuhlukanisa esekelwe ekufundeni kwezibalo, okungelula ukuwela ku-minima yendawo futhi ihlukanise idatha yokuqeqeshwa ngokusebenzisa i-hyperplane elungile ngenkathi idatha yokuqeqeshwa ingase ifakwe kumephu ezicini ezinobukhulu obuphezulu ngokusebenzisa izindlela zemephu ezingaqondile, njengemisebenzi ye-polynomial kanye imisebenzi yesisekelo se-radial. Ngaphezu kwalokho, i-SVM inganikeza i-hyperplane yesinqumo esinembile ngaphansi kwamasampuli alinganiselwe, futhi inekhono elihle lokuhlanganisa. Uma kubhekwa ukusebenza kwayo okuhle kakhulu, i-SVM isetshenziswe kabanzi emikhakheni eminingi. Wang et al. ihlongoze indlela yokuxilongwa kwephutha ehlakaniphile esekelwe ekuhlanganisweni kwe-permutation entropy enezilinganiso eziningi ezihlanganisiwe ezijwayelekile (GCMWPE) kanye ne-SVM [7], engakhipha izici zokuthela ezikalini eziningi ukuze kwakhiwe iqoqo lesici sobukhulu obuphezulu. Bayati et al. uphakamise indlela yendawo enephutha ye-DC microgrid esekelwe ku-SVM [8]. Ngokusebenzisa inani elilinganiselwe lendawo ekugcineni komugqa ngamunye, indawo enembile yephutha eliphezulu le-impedance ingatholakala, futhi imiphumela yokuhlola ibonisa ukuthi uhlelo luqinile emsindweni nokunye ukuphazamiseka. Ref. [9] uphakamise indlela ehlakaniphile yokuxilongwa kwephutha yebhethri ye-lithium-ion esekelwe emshinini we-vector osekelayo, osebenzisa ukuhlunga okucacile kwe-cosine ukuze kuqedwe umsindo.