IoT -Enenziwe i-AI Dry Cooler ye-Hyperscale Data Centres
Kumaziko amakhulu - ombane, apho ukugcina amaqondo obushushu awona aphezulu kwisixhobo esibalulekileyo kubaluleke kakhulu, i-cooler eyomileyo eyenzelwe iimfuno zokusebenza eziphezulu - ivelele. Isakhono sayo sokuphatha ubushushu obuphezulu, obufikelela kwi-100 kW/m², buyenza ibe yi-asethi ebalulekileyo kuncedo lokuvelisa amandla. Uxinaniso oluphezulu - olugciniweyo - sisitshintshisi sobushushu setyhubhu, esine-12 fins nge-intshi nganye, sifikelela kumlinganiselo ochukumisayo wokudlulisa ubushushu obuyi-800 W/m²K, iqinisekisa ukulahla ubushushu okusebenzayo.
Esi sipholileyo somileyo sisebenza kulwelo - olupholileyo, oluvaliweyo - lwenkqubo yelophu enamanzi adiyoni. Kwimizi-mveliso yamandla, olu yilo lunciphisa imingcipheko yokuqhuba umbane, ikhusela amacandelo ombane abuthathaka kunye neenkqubo. Ipholisa ngokufanelekileyo i-coolant, inciphisa ubushushu bayo ukusuka kwi-25℃ukuya kwi-15 degree, ebaluleke kakhulu ekusebenzeni okuzinzile kwezixhobo zamandla ezahlukeneyo - eziphezulu, ezifana nendlela exhasa ngayo ii-GPU kunye namaqela e-CPU kwiindawo ze-HPC.
Ulungelelwaniso lwefeni olungafunekiyo lunikezela ngokusilela - ukusebenza ngokukhuselekileyo, ukuthintela ukuvalwa okungalindelekanga okunokuphazamisa ukuveliswa kwamandla. Ukongeza, i-acoustic enclosure inciphisa ingxolo ukuya kwi-55 dB (A), idala indawo yokusebenza ekhululekile kubasebenzi bezityalo. Elungiselelwe ubukhulu bezityalo - zamandla ombane, le cooler eyomileyo idlala indima ebalulekileyo ekugcineni ukuthembeka kunye nokusebenza kakuhle kwamandla - kweenkqubo zokuvelisa, njengokubaluleka kwayo kuqeqesho lwe-AI kunye nokulinganisa kwesayensi ngaphakathi kwamaqela e-HPC.
Yenzelwe i-hyperscale cloud computing, le IoT{0}}idityanisiwe ye-AI Dry Cooler idibanisa ugcino oluqikelelwayo kunye{1}}nokwenyani kokusebenza kwexesha. Ixhasa ukusebenza kwe-24/7 kumaziko edatha, amaqela e-HPC, kunye -nezityalo ezinkulu zamandla, ukusingatha ukushisa kobushushu ukuya kwi-100 kW/m² ngelixa ugcina i-PUE <1.15.
Iimpawu eziphambili
|
Ipharamitha |
Iinkcukacha |
|
Ukuphatha ubushushu be-Flux |
100 kW/m² (ilungele{1}}iirekhi zeseva ezixineneyo) |
|
Lugqityiwe{0}Uyilo lweTubhu |
12 amaphiko nge-intshi nganye, 800 W/m²·K umlinganiselo wokhutshelo lobushushu |
|
Ukupholisa okuPhakathi |
Amanzi adiyiniweyo (avaliweyo{0}}ukupholisa ulwelo oluludwe lwelophu) |
|
Uluhlu lobushushu |
Indawo yokungena: 25 idigri → Indawo yokuphuma: 15℃(±0.5℃echanekileyo) |
|
Ukudityaniswa kwe-IoT |
Uqhagamshelwano lwe-4G/LTE, ideshibhodi yohlalutyo lwe-AI |
|
Inqanaba Lengxolo |
55 dB(A) (indawo ebiyelweyo yeacoustic ibandakanyiwe) |
Iinkcukacha zobuGcisa
IoT kunye ne-AI yezakhono
◆ Predictive Maintenance: AI algorithms ihlalutya idatha yokwenyani{0}}yexesha (isantya sefeni, ukuhluka kobushushu, ukungcangcazela) ukuqikelela ukusilela kwiiyure ezingama-72 kwangaphambili.
◆ Ukugcinwa kwedatha ngokwembali kuvumela uhlalutyo lwentsingiselo yeshedyuli yolondolozo olusebenzayo.
◆ Ukujongwa okukude: I-app yeselula ivumela{0}}ukulandelela ixesha lokwenyani le-PUE, amazinga okuqukuqela okupholisa, kunye nokusetyenziswa kwamandla. Izilumkiso zibangela ngokuzenzekelayo izinto ezingaqhelekanga (umzekelo, izihluzi ezivaliweyo, ukusilela kweefeni).
Ukupholisa ukusebenza
◆ High{0}}Ukutshintshana kobushushu obuxinano: I-12 fins nge-intshi nganye yandisa indawo ephezulu ngelixa inciphisa ukuchasana nokuhamba komoya. Uyilo luxhasa ukulahla ubushushu kwi 500+ iirack zeseva kwiyunithi enye.
◆ I-Redundant Fan System: N+1 ulungelelwaniso lwabalandeli luqinisekisa ukusebenza okuqhubekayo ngexesha lolondolozo. Ifeni nganye ishushu-iyatshintsheka, kufakwe enye<15 minutes.
Ukusebenza kakuhle kwamandla
◆ Eziguquguqukayo-Abalandeli besantya: Iinjini ze-EC zilungelelanisa isantya ngokusekelwe kwi-AI{1}}ingqikelelo yomthwalo oqhutywayo, icutha usetyenziso lwamandla ngama-35% xa kuthelekiswa nemifuziselo{3}}yesantya esisisigxina.
◆ Ugcino lwamandla lonyaka: $8,{1}} ngeyunithi nganye (nge-$0.12/kWh).
◆ I-Deionized Water Loop: Yehlisa iingozi zokuqhutywa kombane, ilungele ukupholisa iinxalenye eziphezulu-zevoltage kwizityalo zamandla kunye namaqela e-HPC GPU.
Iimeko zeSicelo
◆ Amaziko eNkcukacha ye-Hyperscale: Ukupholisa i-10,{1}} iiseva kwiindawo ze-computing yefu, ukugcina amaqondo obushushu aphezulu kwi-AWS/Azure{2}}yemisebenzi yesikali.
◆ I-HPC kunye neZibonelelo zoQeqesho lwe-AI: Ixhasa ukupholisa kwe-GPU arrays kwiilebhu zophando ze-AI (umzekelo, amaqela e-NVIDIA A100), ukuthintela ukubethelwa kwe-thermal ngexesha loqeqesho lokufunda olunzulu.
◆ Izizinzi zoMbane{0}Ezikhulu: Ukupholisa iziguquli kunye neejenereyitha, ukuqinisekisa ukusebenza okuzinzileyo kuncedo lwamandla obushushu, enyukliya, kunye nahlaziyekayo.
ImiGaqo yoRhwebo
◆ Uphawu kunye neSiqinisekiso: [CoolSmart™], CE, ISO 9001, UL Uluhlu
◆ MOQ: Iyunithi ye-1 (ixoxisana ngee-odolo ezininzi)
◆ Amaxabiso:
1-4 iiyunithi: $25,000-$28,000/unithi
iiyunithi 5-10: $22,000-$25,000/unit
Intlawulo: T / T (30% idiphozithi), L / C ekuboneni
Ixesha lokukhokela: Iintsuku ze-15-20 (ezilungiselelwe), iintsuku ezi-7 (iimodeli eziqhelekileyo)
Ukupakishwa: ISPM 15-eqinisekisiweyo iikhreyiti zomthi (2200×1800×2000mm)
Iinzuzo zeMveliso
◆ I-AI{0}}Ubuchule obuqhutywayo: Yehlisa ixesha lokuphumla elingacwangciswanga ngama-90% ngokulungiswa okuxelwe kwangaphambili, ngokuhambelana nemigangatho ye-Uptime Institute Tier III/IV.
◆ I-Modular Scalability: Iiyunithi zinokufaniswa nokusingatha imithwalo yokushisa ukuya kwi-10MW, ilungele ukwandiswa kweziko ledatha.
◆ Iindleko eziphantsi zizonke zoBunini: i-20% yeendleko zokugcinwa eziphantsi xa zithelekiswa nezipholileyo zemveli, kunye newaranti yeminyaka eyi-10 kwizitshintshi zobushushu.
◆ Ukuthotyelwa kwendalo: Idibana noMyalelo we-EU ErP wokonga amandla kunye ne-RoHS kwizithintelo zeziyobisi eziyingozi.
Iifayile ezishisayo: iziko ledatha elomileyo elipholileyo, iziko ledatha laseTshayina abavelisi abapholileyo, ababoneleli, umzi-mveliso
