東海深部地層巖石可鉆性預(yù)測(cè)方法研究
發(fā)布時(shí)間:2021-01-14 08:10
東海深部地層中部分井PDC鉆頭與地層性質(zhì)匹配性不足,導(dǎo)致機(jī)械鉆速低、鉆頭磨損嚴(yán)重等問(wèn)題發(fā)生。以室內(nèi)巖心微鉆實(shí)驗(yàn)數(shù)據(jù)為依據(jù),首先建立了測(cè)井參數(shù)預(yù)測(cè)巖石可鉆性的非線性多元回歸模型,同時(shí)利用多種人工神經(jīng)網(wǎng)絡(luò)方法對(duì)巖石可鉆性進(jìn)行了預(yù)測(cè),結(jié)果表明非線性多元回歸模型預(yù)測(cè)巖石可鉆性與常規(guī)BP神經(jīng)網(wǎng)絡(luò)、級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)、徑向基RBF神經(jīng)網(wǎng)絡(luò)、BP-RBF雙級(jí)聯(lián)神經(jīng)網(wǎng)絡(luò)模型預(yù)測(cè)結(jié)果均具有較高可信度,但BP-RBF雙級(jí)聯(lián)神經(jīng)網(wǎng)絡(luò)模型預(yù)測(cè)效果最好,更適合于東海深部地層巖石可鉆性預(yù)測(cè)。本文研究結(jié)果可為東海深部地層巖石可鉆性預(yù)測(cè)及鉆頭選型提供借鑒。
【文章來(lái)源】:中國(guó)海上油氣. 2020,32(02)北大核心
【文章頁(yè)數(shù)】:8 頁(yè)
【部分圖文】:
圖1_常規(guī)BP神經(jīng)網(wǎng)絡(luò)巖石可鉆性模型示意圖??Fig.?1?Schematic?diagram?of?rock?drillability?grade?value??
?4?491.?98??4.?61??4.?66??0.?98??4.?65??0.?83??14#??4?609.?09??4.?09??4.?15??1.?44??4.?10??0.?18??4?611.?50??4.?18??4.?00??4.?40??4.?19??0.?22??平均相對(duì)誤差??5.?14??2.?52??2.?2級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)可鉆性??為了提高神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)的穩(wěn)定性,考慮在輸人??■與輸出層之間增加連接權(quán)值。級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)??模型姐圏2所示,??圖2級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)巖石可鉆性預(yù)測(cè)模型示意圖??Fig.?2?Schematic?diagram?of?rock?drillability?grade?value??based?on?cascade?BP?neural?network?model??采用級(jí)聯(lián)BP神經(jīng)調(diào)絡(luò)程序?qū)Ρ恚睒颖緮?shù)據(jù)進(jìn)??行學(xué)習(xí),方法同上述常規(guī)BP神經(jīng)網(wǎng)絡(luò)建立方法,建??立一萬(wàn)個(gè)級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)并優(yōu)眩最終優(yōu)選的級(jí)??聯(lián)BP神經(jīng)網(wǎng)絡(luò)的數(shù)學(xué)模型見表4。??利用多元隨狗模型和級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)分別對(duì)??費(fèi)石可鉆性進(jìn)行了興|3襝驗(yàn),其預(yù)測(cè)結(jié)果見表:5。??預(yù)測(cè)的可鉆性級(jí)值與實(shí)#值之間相關(guān)系數(shù)i??=?Q.?981??8,平均相對(duì)誤差為2.?53%,與常規(guī)BP神經(jīng)網(wǎng)絡(luò)接??近,但相比常規(guī)BP神盈網(wǎng)絡(luò),標(biāo)準(zhǔn)誤差=?0.136??7<0.?1643,最大相對(duì)誤差不超過(guò)10%,,殘差平方和??m67¥<£K?674?65穩(wěn)定性相比常規(guī)BP神蓋網(wǎng)??續(xù)有一■隹握升a??2.?3徑向基RBF神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)可鉆性??盡管在神經(jīng)網(wǎng)絡(luò)的實(shí)豚應(yīng)用中,BP神經(jīng)柯絡(luò)占??多數(shù)y但其也有難以克服的局
第32卷第2期??李乾等:東海深部地層巖石可鉆性預(yù)測(cè)方法研究??131??輸出層??輸入層??筆者采用廣義徑向基神經(jīng)網(wǎng)絡(luò)對(duì)表1數(shù)據(jù)進(jìn)行??學(xué)習(xí),設(shè)置誤差容限為0.01,擴(kuò)散因子5,最大神經(jīng)??元個(gè)數(shù)26。構(gòu)建的徑向基神經(jīng)網(wǎng)絡(luò)模型如圖3所??示。建立的數(shù)學(xué)模型見表6。??圖3徑向基RBF神經(jīng)網(wǎng)絡(luò)巖石可鉆性預(yù)測(cè)模型示意圖??Fig.?3?Schematic?diagram?of?rock?drillability?grade?value??based?on?radial?basis?function?neural?network?model??表6徑向基神經(jīng)網(wǎng)絡(luò)巖石可鉆性預(yù)測(cè)模型數(shù)據(jù)??Table?6?Data?of?rock?drill?ability?grade?value?prediction??model?based?on?RBF?radial?basis?function?neural?network??原始數(shù)據(jù)??連接權(quán)重{2,1}??聲波時(shí)差/??密度/??(g*cm-3)??電阻率/??(0*m)??神經(jīng)元??節(jié)點(diǎn)??巖石可??鉆性級(jí)值??79.?75??2.?25??15.?61??1??_?53.?266?4??67.?36??2.?53??40.?14??2??-2.?219?1??65.?36??2.?50??32.?22??3??3.?112?8??63.?85??2.?53??24.?99??4??-1.?942?4??65.?03??2.?52??18.?24??5??1.?457?9??66.?22??2.?55??26.?32??6??5.?227?2??65.?43??
本文編號(hào):2976545
【文章來(lái)源】:中國(guó)海上油氣. 2020,32(02)北大核心
【文章頁(yè)數(shù)】:8 頁(yè)
【部分圖文】:
圖1_常規(guī)BP神經(jīng)網(wǎng)絡(luò)巖石可鉆性模型示意圖??Fig.?1?Schematic?diagram?of?rock?drillability?grade?value??
?4?491.?98??4.?61??4.?66??0.?98??4.?65??0.?83??14#??4?609.?09??4.?09??4.?15??1.?44??4.?10??0.?18??4?611.?50??4.?18??4.?00??4.?40??4.?19??0.?22??平均相對(duì)誤差??5.?14??2.?52??2.?2級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)可鉆性??為了提高神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)的穩(wěn)定性,考慮在輸人??■與輸出層之間增加連接權(quán)值。級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)??模型姐圏2所示,??圖2級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)巖石可鉆性預(yù)測(cè)模型示意圖??Fig.?2?Schematic?diagram?of?rock?drillability?grade?value??based?on?cascade?BP?neural?network?model??采用級(jí)聯(lián)BP神經(jīng)調(diào)絡(luò)程序?qū)Ρ恚睒颖緮?shù)據(jù)進(jìn)??行學(xué)習(xí),方法同上述常規(guī)BP神經(jīng)網(wǎng)絡(luò)建立方法,建??立一萬(wàn)個(gè)級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)并優(yōu)眩最終優(yōu)選的級(jí)??聯(lián)BP神經(jīng)網(wǎng)絡(luò)的數(shù)學(xué)模型見表4。??利用多元隨狗模型和級(jí)聯(lián)BP神經(jīng)網(wǎng)絡(luò)分別對(duì)??費(fèi)石可鉆性進(jìn)行了興|3襝驗(yàn),其預(yù)測(cè)結(jié)果見表:5。??預(yù)測(cè)的可鉆性級(jí)值與實(shí)#值之間相關(guān)系數(shù)i??=?Q.?981??8,平均相對(duì)誤差為2.?53%,與常規(guī)BP神經(jīng)網(wǎng)絡(luò)接??近,但相比常規(guī)BP神盈網(wǎng)絡(luò),標(biāo)準(zhǔn)誤差=?0.136??7<0.?1643,最大相對(duì)誤差不超過(guò)10%,,殘差平方和??m67¥<£K?674?65穩(wěn)定性相比常規(guī)BP神蓋網(wǎng)??續(xù)有一■隹握升a??2.?3徑向基RBF神經(jīng)網(wǎng)絡(luò)預(yù)測(cè)可鉆性??盡管在神經(jīng)網(wǎng)絡(luò)的實(shí)豚應(yīng)用中,BP神經(jīng)柯絡(luò)占??多數(shù)y但其也有難以克服的局
第32卷第2期??李乾等:東海深部地層巖石可鉆性預(yù)測(cè)方法研究??131??輸出層??輸入層??筆者采用廣義徑向基神經(jīng)網(wǎng)絡(luò)對(duì)表1數(shù)據(jù)進(jìn)行??學(xué)習(xí),設(shè)置誤差容限為0.01,擴(kuò)散因子5,最大神經(jīng)??元個(gè)數(shù)26。構(gòu)建的徑向基神經(jīng)網(wǎng)絡(luò)模型如圖3所??示。建立的數(shù)學(xué)模型見表6。??圖3徑向基RBF神經(jīng)網(wǎng)絡(luò)巖石可鉆性預(yù)測(cè)模型示意圖??Fig.?3?Schematic?diagram?of?rock?drillability?grade?value??based?on?radial?basis?function?neural?network?model??表6徑向基神經(jīng)網(wǎng)絡(luò)巖石可鉆性預(yù)測(cè)模型數(shù)據(jù)??Table?6?Data?of?rock?drill?ability?grade?value?prediction??model?based?on?RBF?radial?basis?function?neural?network??原始數(shù)據(jù)??連接權(quán)重{2,1}??聲波時(shí)差/??密度/??(g*cm-3)??電阻率/??(0*m)??神經(jīng)元??節(jié)點(diǎn)??巖石可??鉆性級(jí)值??79.?75??2.?25??15.?61??1??_?53.?266?4??67.?36??2.?53??40.?14??2??-2.?219?1??65.?36??2.?50??32.?22??3??3.?112?8??63.?85??2.?53??24.?99??4??-1.?942?4??65.?03??2.?52??18.?24??5??1.?457?9??66.?22??2.?55??26.?32??6??5.?227?2??65.?43??
本文編號(hào):2976545
本文鏈接:http://sikaile.net/kejilunwen/shiyounenyuanlunwen/2976545.html
最近更新
教材專著