
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Abalobi: 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Shengyu Tao Haizhou Liu Chongbo Sun Haocheng Ji Guanjun Ji Zhiyuan Han Runhua Gao Jun Ma Ruifei Ma Yuou Chen Shiyi Fu Yu Wang Yaojie Sun Yu Rong Xuan Zhang Guangmin Zhou Hongbin Sun Isishweshwe Iibhetri ezingalungisiweyo ezineempahla ezahlukeneyo ze-cathode ziyivimbela inganokwamkelwa ngokuthe ngqo ukugqibelela ngenxa yesimo sabo esithile nge-cathode. Ukwanda kweebhetri ezithotywayo kufuna ukuhlelwa ngokuchanekileyo ukuze kusetyenziswe ngqo ukugqibelela, kodwa imingeni ivela kwiimbali zokusebenza ezahlukeneyo, abavelisi abahlukeneyo, kunye neengxaki zobumfihlo bedatha yamaqabane okugqibelela (abanikazi bedatha). Apha sibonisa, ukusuka kwiseti yedatha eyodwa yeebhetri ezili-130 ze-lithium-ion ezijikeleze izinto ezisi-5 ze-cathode kunye nabavelisi abasi-7, indlela yokufunda yeenjini ezahlukeneyo ingazicwangcisa ezi bhetri zithotywayo ngaphandle kokuxhomekeka kwidatha yokusebenza yangaphambili, ikhusela ubumfihlo bedatha bamalingane okugqibelela. Ngokusebenzisa izinto ezikhethweyo ezivela kumjikelo wokutshaja-ukukhutshwa wokugqibela ubomi, imodeli yethu ibonisa i-1% kunye ne-3% yeempazamo zokuhlela i-cathode phantsi kweemeko zokugqibelela ezifanayo kunye nezahlukeneyo ngokulandelelanayo, ezinxulumene nendlela yethu entsha yokuvotela yobude lobude be-Wasserstein. Ngokwezobuchwepheshe, indlela ecetywayo igxininisa ixabiso lokuhlela ibhetri ngokuchanekileyo kushishino lokugqibelela olunempumelelo nolugcina ubomi. Esi sifundo sibhengeza imodeli entsha yokusebenzisa idatha evakalayo kubumfihlo evela kwimithombo eyahlukeneyo, ilungiselela ukwenza izigqibo ezisebenzisanayo kunye nokuhlonipha ubumfihlo kwiinkqubo ezisasaziweyo. Isingeniso Iibhetri ze-Lithium-ion (LIBs), ezisebenza njengezixhobo zokugcina amandla, zifumene ukusetyenziswa ngokubanzi kwizindawo ezahlukeneyo, ukusuka kwimveliso yamashishini ukuya ebomini bemihla ngemihla njengendlela yobuchwepheshe eyavunyiweyo. Iingxelo zibonisa ukuba ubungakanani bemveliso yehlabathi ye-LIBs bungadlula i-1.3 TWh ngo-2030  xa imfuno ephuphumayo yee bhetri iya kudlula kakhulu kwiindawo zeminyaka eyimbi njenge-lithium kunye ne-cobalt , . Nangona kunjalo, ubomi obuqhelekileyo beemveliso ze-LIB bukhona iminyaka emi-5–8, nto leyo ekhokelela ekwanda okuzayo kweebhetri ezithotywayo kumazwe amaninzi. Ukuba aziphathwa ngokufanelekileyo, iibhetri ezithotywayo ziya kubangela inkunkuma engaqhubekiyo kunye nomonakalo kwindalo. Ngokubhekiselele kwezi meko, ukuphuhliswa kweteknoloji yokugqibelela ibhetri kubaluleke kakhulu njengoko sijamelene namaza ezayo ekuphunyweni kwe-LIB . 1 2 3 4 Ukuphucuka okutsha kuphando lokugqibelela ibhetri kugxile kwizindlela ze-pyrometallurgical, i-hydrometallurgical, kunye ne-direct recycling approaches . Ngokungafaniyo neendlela ze-pyrometallurgical kunye ne-hydrometallurgical, i-direct recycling iyahluka njengendlela eyahlukileyo. Le nkqubo ayibangeli monakalo wesibini kumsebenzi wezinto, evumela ukulungiswa okusebenzayo kunye nokubuyiselwa kokusebenza. Ngaphezu koko, i-direct recycling ibonisa ingeniso ephezulu, njengoko ibandakanya ukusetyenziswa kwamandla okuphantsi, ukukhutshwa kwezegreenhouse okuncitshisiweyo, kunye neenqwelomoya zendalo ezilula , . Kwenenele, kubasebenzi bokugqibelela ibhetri badla ngokudibana ne-LIBs eziqukethe izinto ezingaziwayo okanye iimodyuli zebhetri eziqukethe umxube weentlobo ezahlukeneyo zezinto ze-cathode. Ukucinga ukuba i-direct recycling inokuxhomekeka kakhulu kwi-cathode, ubunzima obunjalo benza ukusetyenziswa kwe-direct recycling kungasebenzi ukuze kuzuze ixabiso kweebhetri ezithotywayo . Kubalulekile ukugxininisa ukuba nokuba izimbiwa ezibalulekileyo ezivela kwiintlobo zemixube ye-cathode zingafunyanwa kusetyenziswa iindlela zokugqibelela eziqhelekileyo, ukusebenzisana kwezinto ezahlukeneyo ze-cathode ngexesha lenkqubo yokugqibelela kungonakalisa umgangatho weemveliso . Ngoko ke, ukuqonda ulwazi lwentlobo yezinto ze-cathode ngakwicala lokugqibelela kuchaphazela kakhulu ukhetho lwendlela ye-direct recycling kwaye ekugqibeleni kuphucula umgangatho weemveliso, ingeniso, kunye nokugcinwa. 5 6 7 8 9 Ukugqibelela okusizwa ngabantu kuye kwacetyiswa ukuba kuchonge ulwazi lwentlobo yezinto ze-cathode ezithotywayo kwinqanaba lokulungiswa kwangaphambili, nto leyo engazange ibe nenzuzo ngokwezobuchwepheshe xa ishishini lokugqibelela likhula . Ukufumana ngokufanelekileyo ulwazi lwentlobo ye-cathode etotywayo, uluntu olungokwesayensi kunye nolweshishini luye lwazisa inkqubo yokulandelela ubomi bebhetri  kunye neengcamango ezivelayo ezifana nephepha lebhetri  kunye ne-genome yedatha yebhetri . Nangona iibhetri ezininzi ziye zasetyenziswa phambi kwezi zinyathelo, kukho ingqibelelo yokwanda yokuba ulwazi lwebhetri kufanele lufumaneke kulo lonke ithuba lokugcina izigqibo zokubuyiselwa kwesibini . Oku kubonakala ngakumbi kwi-sector yokugqibelela ibhetri, inqanaba lokugqibela lobomi besibini bebhetri, njengoko indlela yokugqibelela ingaba ne-cathode ethile. Nangona kunjalo, iinkqubo zokulandelela ubomi bebhetri okanye iipassport zebhetri zinikwa amandla zizixhobo zombane ezifana neebhodi ezibhatalayo kunye neenkqubo zokunxibelelana ezikufuphi, ezingaletha utyalo-mali oluninzi kwaye zingahambelani neeyunithi ezahlukeneyo zoyilo lwebhetri. Ngaphezu koko, izixhobo zombane zihlala zinzima ukuziphatha ngokungaguquguqukiyo kubomi bazo bonke, nto leyo ekhokelela kwizixhobo ezigugileyo kunye nokungafumaneki kwinqanaba lokugqibelela njengoko inkqubo yanamhlanje yokuvelisa i-LIBs ingekabikho kumgangatho wokwenza-kugqibelela . Ngoko ke, ukuphuhliswa okungaphezulu kuyimfuneko ukufezekisa ukuhlelwa okusebenzayo kwenkqubo ye-cathode ibhetri kusetyenziswa ulwazi lwenkundla olufikeleleka ngokulula , , ngokungafaniyo nedatha yembali egciniweyo okanye ngendlela esiziswa ngabantu, ilungiselela ukwamkelwa kokugqibelela okuthe ngqo ukuphucula umgangatho kunye nenzuzo yezinto ezigqityiweyo. 1 10 11 12 13 14 15 16 Kukho iminyaka embalwa, ukufunda ngomatshini kuye kwavela njengegajethi eluncedo yokujongana nemibuzo evulekileyo kuzo zonke iindawo zebhetri. Kwezinye izihloko ezinxulumene nebhetri, ukufunda ngomatshini kuye kwavumela ukufunyanwa ngokuzenzekelayo kweendlela ezintsonkothileyo zebhetri , , , ukubikezela ubomi obusebenzayo obushiyekileyo , , , , , ukuxabisa imeko yempilo , , , ukulungelelanisa iprofayili yokujikeleza , , ukubikezela ukwabiwa kobomi , nokude ukukhokela uyilo lwebhetri , , kunye nokubikezela ukusebenza kobomi bonke kwangoko emva kokwenziwa . Kwimeko yokugqibelela ibhetri, imisebenzi embalwa kuye yaphonononga ukufunda ngomatshini ngokubhekiselele kwii-athomu ze-cathode , , nto leyo ixhomekeke kwidatha enqabileyo yebhetri, ingakumbi kwabo bagqityiweyo kwisigaba sokugqibela ubomi. Uninzi lwezifundo ezipapashiweyo zibonisa ubungakanani besampula obuncinci kakhulu  kwaye bunqongophala ngakumbi kubumelelo bebhetri . Ubunqongophala bubangelwa yindleko enkulu, ixesha elide lokuvavanya , kwaye, okona kubaluleke kakhulu, ubumfihlo bedatha ngenxa yeengxaki zorhwebo okanye izilangazelelo. Ngokulandelelana, ingxaki yobumfihlo yenza idilema apho idatha ebhetri ekhoyo, nangona inkulu ngobungakanani kunye nobume obwahlukeneyo ukusuka kumacandelo amaninzi afana nabavelisi beebhetri, izicelo zokusebenza, imibutho yemfundo, kunye neenqanawa zeqela lesithathu, ayinakwabelwana ngayo. Idilema enjalo ibiza ukufunda ukuhlelwa kwe-athomu ye-cathode ukuze kulungiswe ukukhetha indlela yokugqibelela ibhetri ngendlela esebenzisanayo ngelixa ihlonipha ubumfihlo. 17 18 19 20 21 22 23 24 19 25 26 27 28 29 30 31 32 33 34 35 36 37 I-federated machine learning, njengeparadigm echanekileyo kunye nekhusela ubumfihlo, inamandla okusombulula zombini ukwenza intsebenziswano yamacandelo amaninzi (ekulinganayo, ubungakanani bedatha yebhetri kunye nohluko) kunye neengxaki zobumfihlo ngokusebenzisa ukufunda ngomatshini ngokusebenzisana , , . Kulo lonke ixesha lokufundisa, abanini bedatha abasasaziweyo benza uqeqesho lwasekuhlaleni ngamandla abo okubala apho, bacofa imiphumo/iziphumo zemodeli eqeqeshwe ngolwazi, kwaye bayithumele kumxibeleli ongaphambili ukuze kuqwalaselwe. Amaxwebhu okuba iiseti zedatha ekrwada azikaze ziphume kubanini bazo bedatha kwaye iiparameters/iziphumo ezidluliselweyo zicofwe ngokufanelekileyo ukukhusela ubumfihlo bedatha. I-federated machine learning iye yaphononongwa ngokubanzi kumacandelo amaninzi okusebenza, abandakanya impilo kawonkewonke , , ukuxilongwa kwezonyango , , , i-e-commerce , i-Internet of Things , i-mobile computing , kunye ne-smart grid , , . Le ndlela ingaguqula i-paradigm yophando enikwe yi-data kwizona ndawo zamandla ngokuvumela ukwenza intsebenziswano ekhusela ubumfihlo, ingakumbi kwabo banokufikelela kwidatha encinci. Malunga ne-sector yokugqibelela ibhetri, i-federated machine learning iyathweswa iziqu ezithembisayo ekusebenziseni ubungakanani obukhulu bedatha yebhetri ekhoyo kodwa engenakwabelwana ngayo ngenxa yeengxaki zobumfihlo. Ngale paradigm esebenzisana ngelixa ikhusela ubumfihlo, ukuhlelwa kwebhetri etotywayo kungenziwa ngokuchanekileyo okuphezulu, ukusebenza kakuhle, ukwanda, kunye nokugcinwa, kulungiselela umgangatho kunye nenzuzo yezinto ezibuyisiweyo. Ngokwazi kwethu, izifundo ze-federated machine learning ezigxile ekugqibeleleni ibhetri azikaze ziphinde zibhaliswe. 38 39 40 41 42 43 44 45 46 47 48 49 50 51 Kolu phando, senze ukuhlelwa kwe-athomu yezinto ezithotywayo, sisebenzisa idatha ebhetri ekhoyo ukusuka kumalingane amaninzi, afana nabavelisi beebhetri, abaqhubi besicelo sokusebenza, imibutho yophando lwezemfundo, kunye neenqanawa zeqela lesithathu, ngendlela yokufunda ngomatshini esebenzisanayo kodwa ekhusela ubumfihlo njengoko ibonisiwe kwiFig. 1. Imodeli yethu ye-federated machine learning yafundiswa kusetyenziswa umjikelo omnye kuphela wedatha yovavanyo lwenkundla ngegajethi yobutyebi obumiselweyo, ngaphandle kolwazi lwangaphambili lweemeko zokusebenza zembali. Siyathelekisa amandla okubikezela emodeli yethu ye-federated machine learning kunye neyemodeli zasekuhlaleni ezifundiswe ngokuzimeleyo esekwe kwidatha yasekhaya phantsi kweemeko zokugqibelela ibhetri ezifanayo nezahlukeneyo. Ingxaki yokwahluka isonjululwe yindlela yethu ecetywayo yokuvotela yobude lobude be-Wasserstein. Ukuvavanywa kwezobuchwepheshe kokugqibelela ibhetri etotywayo kusetyenziswa ifremu yethu ye-federated machine learning iyenziwa, igxininisa ukubaluleka kunye nemfuneko yokuhlela ibhetri etotywayo ngokuchanekileyo. Sikhuphela ngokucacileyo ukuchazwa kwemodeli, iziphumo zokugqibelela ibhetri, kunye neendawo ezibanzi zemisebenzi ezayo yokugqibelela ibhetri edibaniswe ne-federated machine learning.    Imithombo yedatha emininzi, efana nabavelisi beebhetri (Umnikelo womfanekiso uvela kwi-Addionics), abaqhubi besicelo sokusebenza (ibhhetri ephuma kwinqanaba lombane lemoto kaTesla. Umnikelo womfanekiso uvela kwiTesla), imibutho yophando lwezemfundo, kunye neenqanawa zeqela lesithathu, zingaba ngabanikeli bedatha. Idatha yebhetri ayitshintshiswa phakathi kwabanikeli okanye ikhokhelwe kumqhubi webhetri. Endaweni yoko, abanikezeli bedatha baqeqesha iimodeli zasekuhlaleni kwaye babelane ngeeparameter zemodeli kunye nomqhubi webhetri ukwakha imodeli yehlabathi. Indlela yokuvotela yobude lobude be-Wasserstein ecetywayo idibanisa iimodeli zasekuhlaleni kwimodeli yehlabathi, ezinokwenza umonakalo ukungalingani kwedatha kunye nengxolo. Abaqhubi beebhetri bangasebenzisa imodeli eyakhiwe ngokudibeneyo yokuhlela ibhetri, idibene nedatha yovavanyo lwenkundla enokufikeleleka ngokulula.   Ifremu yethu ye-federated machine learning iyakhuthaza abalingane ukuba babelane ngedatha ngelixa kugcinwa ubumfihlo bedatha ngokuchasene nepharadigim yemveli yokuhlukanisa idatha. a b Iziphumo Ukuqokelelwa kwedatha kunye nokumiselwa Iikinetics ezahlukeneyo zebhetri kwiintlobo zebhetri ezahlukeneyo ziphezulu kakhulu kwaye zinzima ukuchazwa ngenxa yeemeko zokusebenza ezahlukeneyo, ukwahluka kokuveliswa, kunye nokusetyenziswa kwembali . Ukufumana isisombululo kule dilema, siye saqokelela kwaye samiswa iibhetri ezithotywayo ezili-130 ezineentlobo ezisi-5 ze-cathode ezivela kubavelisi abasi-7 ukwakha idatha engaphandle kokusasazeka, ngokulinganayo idatha eyahlukeneyo. Kuthathwa iimbali zokusetyenziswa ezahlukeneyo, amandla eebhetri eziqokelelweyo angaphantsi kwe-90% yamandla ayo. Izinto ze-cathode zebhetri ziyi-lithium cobalt oxide (LCO), i-nickel manganese cobalt (NMC), i-lithium ferrophosphate (LFP), i-nickel-cobalt-aluminum oxide (NCA), kunye nemixube ye-NMC-LCO, eyayongezwa kwizigaba ezisi-9 ngokusekelwe kubavelisi (Itheyibhile yesiXhasi ). Siyifake ngabom iibhetri ezineembali ezahlukeneyo zokusetyenziswa, ukusuka kuvavanyo lwelebhu ukuya kwiipati zokudlala zemoto yombane, ukuqeqesha imodeli enokugcinwa yomqhubi webhetri ngaphandle kolwazi lwembali kunye neentlobo zebhetri. 52 1 Ngokumiselwa, yonke idatha efumaneka kumqhubi yeyangoku (yovavanyo lwenkundla) umjikelo ngovavanyo olunye lokutshaja kunye nokukhutshwa, olulula ukulwenza kwiimeko zokusebenza. Idatha ekhoyo ngoku ivunyiwe ngokucoca amanqaku alahlekileyo, ukubuyisela iziphumo, kunye nokwenza uhlalutyo lwe-median. Ingxolo ebangelwa ngumntu kunye neyobumeko bebhetri igcinwe ngabom, nangona kunjalo, ukwenza imodeli inokwenza umonakalo kwiimfuno ezingagqibelelanga. Idatha ke iye yahlazwa ngokuthe ngqo ukuze kugqityezelwe umgca (Umboniso owongezelelweyo ) kwaye yahlolwa ukuze kuncitshiswe ubungakanani, kunye neseti ezifanayo yeenjini zokumiselwa (Inqaku elongezelelweyo ). Iinjini ezikhethwe kwi-pipeline yokumiselwa ziyaqondakala kakhulu, ingxaki enomdla omkhulu wezorhwebo. Ngokwazi kwethu, yeyona nto yokuqala idatha eyahlukeneyo yebhetri esuka kumithombo emininzi kunye nembali yokuphila esetyenziselwa ukuncedisa uyilo lwesicwangciso sokugqibelela ibhetri. 1 1 Umboniso 2a, 2b ubonisa inkqubo yenjini. Sigxile kumgca wokutshaja kunye nokukhutshwa kwebhetri etotywayo kumjikelo wokugqibela, oko kukuthi, umjikelo omnye wokutshaja kunye nomnye wokukhutshwa (Imiboniso eyongezelelweyo 2–5). Kwijikelo lokutshaja, izinto ezili-15 zifunyanwa kumgca we-voltage-capacity kunye ne-dQ/dV curve, apho i-V kunye ne-Q zichaza amaxabiso voltage kunye namandla, ngokulandelelanayo. Iseti efanayo yeenjini zifunyanwa kumjikelo wokukhutshwa. Ngenxa yoko, izinto ezingama-30 zifunyanwa ngoku, njengoko kukhonjisiwe kwi-F1 ukuya kwi-F30. Jonga kwiThebula elongezelelweyo 2 kunye neNqaku eliLonyezelweyo 2 ukuze ufumane inkcazo eneenkcukacha zezinto. Umboniso 2c ubonisa ubungakanani obupheleleyo kunye nobunxulumene nezinto ezikhethiweyo zebhetri ukusuka kwiklasa nganye. Uninzi lwamaxabiso enjini aphakathi kwezi klasi lugqumana kwindawo ye-−1 ukuya kwi-0 (nombala oluhlaza okukhanyayo) kwaye alunakuqhagamshelwa, lubonisa ubunzima bokucwangcisa uhlobo lwebhetri kusetyenziswa umjikelo omnye wedatha yebhetri. Ubunzima bulindeleke kuba iimeko zangaphambili zokusetyenziswa ezingaqhelekanga zingachaphazela i-kinetics yokutshaja-ukukhutshwa kwee bhetri ukuze izinto ezikhethiweyo zibe nomdla omkhulu nangona iintlobo zebhetri ezahlukeneyo (Umboniso owongezelelweyo 6). Endaweni yokutolika ngokut

Part of HackerNoon's growing list of open-source research papers, promoting free access to academic material.

Le audio iveliswe ngolwimi lokuqala lwebali!

Ezi AI leyi Tiyisaka uBumfihlo Ingaenza Imali Enkulu Ngemiyalezo yeebhetri

About Author

IZIMVO

ZIJONGE IIMPAWU

ELI NQAKU LINIKEZELWE KU

Related Stories

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps