朱海濤 林伯韜 石蘭香 竇升軍
摘要:為提高油藏注水開(kāi)發(fā)效率,制定精細(xì)的分層配產(chǎn)配注制度,須針對(duì)水平井開(kāi)展吸水及產(chǎn)液剖面的高效檢測(cè)。對(duì)此,提出基于自適應(yīng)矩估計(jì)優(yōu)化算法(Adam算法)的水平井吸水及產(chǎn)液剖面測(cè)溫反演方法。該方法首先利用儲(chǔ)層與井筒內(nèi)的溫度分布模型構(gòu)建反演目標(biāo)函數(shù),其次通過(guò)Adam優(yōu)化算法,在最優(yōu)化反演目標(biāo)函數(shù)的過(guò)程中定量獲取流動(dòng)剖面。將該方法應(yīng)用于阿曼Safah油田及新疆風(fēng)城油田的兩口水平井,采用生產(chǎn)測(cè)井工具測(cè)得各井段吸水量和井口測(cè)量的產(chǎn)液量對(duì)反演結(jié)果進(jìn)行驗(yàn)證。結(jié)果表明:建立的反演方法不需要求解復(fù)雜的耦合模型,計(jì)算效率高,不僅可以定量監(jiān)測(cè)流動(dòng)剖面的動(dòng)態(tài)變化、評(píng)價(jià)各層段的貢獻(xiàn)率,還可半定量刻畫水平井各層段相對(duì)滲透率的演化規(guī)律,指導(dǎo)現(xiàn)場(chǎng)制定更加精細(xì)化的配產(chǎn)、配注及增產(chǎn)方案。
關(guān)鍵詞:吸水剖面; 產(chǎn)液剖面; 分層配產(chǎn)配注; 溫度反演; Adam優(yōu)化算法
中圖分類號(hào):TE 357 文獻(xiàn)標(biāo)志碼:A
引用格式:朱海濤,林伯韜,石蘭香,等.基于Adam優(yōu)化算法的水平井流動(dòng)剖面測(cè)溫反演方法[J].中國(guó)石油大學(xué)學(xué)報(bào)(自然科學(xué)版),2023,47(2):99-107.
ZHU Haitao, LIN Botao, SHI Lanxiang, et al. An inversion method to calculate horizontal well flow profile using temperature data based on Adam optimization algorithm[J].Journal of China University of Petroleum(Edition of Natural Science),2023,47(2):99-107.
An inversion method to calculate horizontal well flow profile using temperature data based on Adam optimization algorithm
ZHU Haitao1, LIN Botao1, SHI Lanxiang2, DOU Shengjun3
(1.College of Artificial Intelligence, China University of Petroleum(Beijing), Beijing 102249, China; 2.Research Institute of Petroleum Exploration & Development, PetroChina, Beijing 100083, China;3.Xinjiang Huishi Petroleum Technology Company Limited, Karamay 834000, China)
Abstract: In order to improve the efficiency of water injection and design a finely stratified production and injection workflow, an efficient detection of the injection and production flow profile along a horizontal well is required. In this regard, a method based on the adaptive moment estimation optimization (Adam) algorithm was proposed to inverse the injection and production profile in a horizontal well using the temperature data. First, a temperature distribution model in both the reservoir and wellbore was applied to construct an inversion target function. Secondly, the Adam optimization algorithm was used to quantitatively obtain the fluid flow profile from solving the optimization target function. The method was used to analyze the flow profile of two horizontal wells in Safah Oilfield (Oman) and Fengcheng Oilfield (Xinjiang, China), respectively. The inversion results were verified against the sectional injectivity measured by production logging down hole and the total production measured on the surface. The results show that the inversion method has high computational efficiency without solving complex coupling models. It is not only capable of quantitatively monitoring the dynamic evolution of the flow profile and evaluating the contribution rate of each layer section, but also provides a semi-quantitative portrayal of the evolution law for the relative permeability in each section of horizontal wells, guiding the field engineers to design a more finely production and injection workflow, and stimulation schemes.
Keywords: injectivity profile; production profile; stratified production and injection; inversion of temperature data; Adam optimization algorithm
注水井的吸水剖面和生產(chǎn)井的產(chǎn)液剖面資料是進(jìn)行油田開(kāi)發(fā)動(dòng)態(tài)分析必不可少的依據(jù)[1-3]。目前,水平井中吸水和產(chǎn)液剖面的監(jiān)測(cè)主要通過(guò)生產(chǎn)測(cè)井工具(PLT)實(shí)現(xiàn)。傳統(tǒng)的PLT在水平井中應(yīng)用時(shí)存在輸送困難、容易卡井、成本高且不能勝任長(zhǎng)期監(jiān)測(cè)任務(wù)的問(wèn)題[4-6]。相較于PLT作業(yè),水平井筒中的溫度測(cè)量相對(duì)簡(jiǎn)便快捷。例如分布式光纖溫度傳感系統(tǒng)(DTS)可靈活安裝在套管內(nèi)外,提供高精度的井筒溫度數(shù)據(jù)[7]。利用油藏和井筒之間的傳熱傳質(zhì)規(guī)律,國(guó)內(nèi)外一些學(xué)者建立了相關(guān)的理論分析模型,將溫度數(shù)據(jù)解釋為相應(yīng)的流動(dòng)剖面。Hashish等[4,8-9]提出了基于傳熱模型解析解曲線圖版的吸水剖面反演方法。Pimenov等[10]介紹了基于最小二乘算法的水平井吸水剖面溫度反演方法。
Zhang等[11]、Yoshioka等[12-14]、Sui等[15]、朱世琰[16]、蔡珺君等[17-18]建立了基于萊溫伯格-馬夸特算法(L-M算法)的水平井產(chǎn)液剖面溫度反演方法。Li等[19-20]、Luo等[21-27]分別建立了基于L-M算法、馬爾可夫鏈蒙特卡羅算法(MCMC 算法)以及模擬退火算法(SA算法)的水平井產(chǎn)出剖面溫度反演方法。然而,上述的研究多是單獨(dú)針對(duì)吸水或產(chǎn)液剖面的反演方法,對(duì)流動(dòng)剖面通用性反演方法的探討較為欠缺,且相關(guān)的反演方法需要求解復(fù)雜的耦合模型;當(dāng)反演模型復(fù)雜、參數(shù)較多時(shí),迭代次數(shù)顯著增加,反演效率低,模型收斂慢,容易出現(xiàn)不收斂的情況。為了能長(zhǎng)期實(shí)時(shí)監(jiān)測(cè)水平井的流動(dòng)剖面,筆者提出基于可自動(dòng)調(diào)節(jié)步長(zhǎng)的自適應(yīng)矩估計(jì)優(yōu)化算法(Adam算法)建立水平井流動(dòng)剖面測(cè)溫反演方法,并將此方法應(yīng)用于油田現(xiàn)場(chǎng),利用現(xiàn)場(chǎng)測(cè)量數(shù)據(jù)對(duì)反演結(jié)果進(jìn)行驗(yàn)證,并分析水平井注入及生產(chǎn)過(guò)程中的流動(dòng)剖面及各層段相對(duì)滲透率的演變過(guò)程。
1 流動(dòng)剖面反演方法
基于能量守恒定律,首先建立水平井非等溫流體注入及產(chǎn)出過(guò)程中瞬態(tài)溫度分布函數(shù),其中包括儲(chǔ)層溫度分布函數(shù)和井筒溫度分布函數(shù)。反演方法主要包括反演目標(biāo)函數(shù)的構(gòu)建和反演目標(biāo)函數(shù)的最優(yōu)化。反演目標(biāo)函數(shù)表示為儲(chǔ)層溫度分布函數(shù)與井筒溫度分布函數(shù)在井壁附近的誤差函數(shù)。鑒于Adam算法具有自適應(yīng)調(diào)節(jié)步長(zhǎng)(學(xué)習(xí)率)和參數(shù)更新效率高的優(yōu)點(diǎn),采用該算法作為反演目標(biāo)函數(shù)的優(yōu)化算法。將反演目標(biāo)函數(shù)作為Adam優(yōu)化算法中的目標(biāo)函數(shù),流動(dòng)剖面作為反演目標(biāo)參數(shù),在目標(biāo)函數(shù)最優(yōu)化過(guò)程中,定量確定反演目標(biāo)參數(shù)(流動(dòng)剖面)。
1.1 儲(chǔ)層溫度分布函數(shù)
假設(shè)箱型油藏中有一口水平井,取半徑差為dr的微元,考慮流體在儲(chǔ)層滲流過(guò)程中的熱傳導(dǎo)和熱對(duì)流,忽略黏性耗散和熱膨脹等微量熱效應(yīng)對(duì)儲(chǔ)層溫度剖面的影響。在注液及產(chǎn)液條件下微元中熱能傳輸如圖1所示。注入過(guò)程中進(jìn)入微元中的熱量為流體攜帶的熱量q(x,r,t)和微元前緣通過(guò)熱傳導(dǎo)進(jìn)入微元的熱量qT(x,r+dr,t);流出微元的熱量包括流體從微元前緣流出時(shí)攜帶的熱量q(x,r+dr,t)和在r處通過(guò)熱傳導(dǎo)流出的熱量qT(x,r,t)。產(chǎn)液過(guò)程中微元中的熱量傳輸僅方向與注液過(guò)程中不同,如圖1所示。基于能量守恒方程,注入及產(chǎn)液過(guò)程中水平井的井周儲(chǔ)層徑向溫度分布為
中,ρr和ρf分別為儲(chǔ)層巖石和流體的密度,kg/m3;cr和cf分別為儲(chǔ)層巖石和流體的比熱容,J/(kg·K);φ為巖石孔隙度;λr和λf分別為儲(chǔ)層巖石和流體的導(dǎo)熱系數(shù),W/(m·K);um為儲(chǔ)層中流體的體積流量,m3/s;T為儲(chǔ)層溫度,K。
式(1)中等號(hào)右邊項(xiàng)為熱傳導(dǎo)項(xiàng),等號(hào)左邊第一項(xiàng)為微元中的熱能變化量,第二項(xiàng)為熱對(duì)流項(xiàng)。注水過(guò)程與產(chǎn)液過(guò)程中儲(chǔ)層溫度分布方程的區(qū)別在于熱對(duì)流項(xiàng)的正負(fù)號(hào),這是由于注液及產(chǎn)液過(guò)程中流體流動(dòng)方向相反。流體在儲(chǔ)層中的滲流速度u為
(3)通過(guò)Adam算法優(yōu)化反演目標(biāo)函數(shù)。將注液和產(chǎn)液工況下的反演目標(biāo)函數(shù)(式(18)和(19))分別作為Adam算法的目標(biāo)函數(shù),對(duì)其進(jìn)行優(yōu)化,使反演目標(biāo)函數(shù)滿足約束條件。具體的反演流程如圖4所示。其中初始參數(shù)包括步長(zhǎng)(學(xué)習(xí)率)α,指數(shù)衰減率β1和β2;k為時(shí)間步長(zhǎng),初始值為0;f(k)為目標(biāo)函數(shù)在時(shí)間步長(zhǎng)k處的梯度;m(k)、v(k)分別為在時(shí)間步長(zhǎng)k處的偏一階矩估計(jì)和偏二階矩估計(jì);(k)、(k)分別為在時(shí)間步長(zhǎng)k處的修正偏一階矩估計(jì)和修正偏二階矩估計(jì);誤差限e=10-10。
(4)當(dāng)Adam算法優(yōu)化完畢時(shí)判斷S(g)<ε是否成立。若成立則g*=g(k+1),即為所求的各離散區(qū)間內(nèi)的流動(dòng)剖面;若不成立,則調(diào)整參數(shù),繼續(xù)使用Adam算法計(jì)算,直至S(g)<ε成立。
3 現(xiàn)場(chǎng)應(yīng)用
3.1 實(shí)例井概況
實(shí)例井1為位于阿曼Safah油田的某口水平注水試驗(yàn)井FS-1,垂深1 420 m,水平段長(zhǎng)118 m,其中包括水平段開(kāi)始處的一個(gè)長(zhǎng)度為28 m的非滲透段(套管段)。注入流體為水,注入時(shí)間為4 h,井徑為177.8 mm,注入流量為150 m3/d。采用熱電偶測(cè)量水平井筒中不同位置的溫度。水平段開(kāi)始時(shí)的進(jìn)口流體溫度為40 ℃,原始地層溫度為70 ℃。其中FS-1井的地層導(dǎo)熱系數(shù)為3 W/(m·K);地層比熱容為1 000 J/(kg·K);地層密度為2 600 kg/m3;注入液的比熱容為4 127 J/(kg·K);注入液的導(dǎo)熱系數(shù)0.67 W/(m·K);地層初始?jí)毫?0.5? MPa。熱電偶測(cè)得水平段各點(diǎn)的溫度數(shù)據(jù)如圖5所示[10]。其中溫度梯度較大的兩個(gè)位置分別處于套管非滲透段和井筒末端。
實(shí)例井2是位于新疆風(fēng)城油田重32區(qū)某SAGD井組中的產(chǎn)液井FS-2,垂深200 m,水平段長(zhǎng)413 m。此SAGD井組于2009年投產(chǎn),蒸汽腔發(fā)育穩(wěn)定。在生產(chǎn)過(guò)程中注汽井恒壓注入熱蒸汽,對(duì)產(chǎn)液井來(lái)說(shuō)蒸汽腔可近似看作穩(wěn)態(tài)邊界。FS-2井的井身結(jié)構(gòu)及溫度測(cè)量位置如圖6所示,現(xiàn)場(chǎng)采用熱電偶測(cè)量水平井筒中離散點(diǎn)的溫度。FS-2井的原始地層溫度為18 ℃,其上方蒸汽腔中的溫度約為233 ℃。FS-2井的地層導(dǎo)熱系數(shù)為3.15 W/(m·K);地層比熱容為890 J/(kg·K);地層密度為2 600 kg/m3;產(chǎn)出液的比熱容為3 200 J/(kg·K);井徑為177.8 mm;地層初始?jí)毫?.8 MPa?,F(xiàn)場(chǎng)分別測(cè)量了FS-2井在5個(gè)不同時(shí)間點(diǎn)(間隔30 d)的溫度數(shù)據(jù),即從測(cè)溫開(kāi)始第30、60、90、120和150 d的數(shù)據(jù),分為FS-2-1~FS-2-5共5組數(shù)據(jù),如圖7所示。
3.2 吸水剖面反演
將FS-1井的相關(guān)參數(shù)及溫度測(cè)量數(shù)據(jù)代入反演目標(biāo)函數(shù)(式(18)),按照流程進(jìn)行反演,其中反演誤差ε設(shè)置為10-4。當(dāng)反演誤差函數(shù)滿足預(yù)設(shè)精度后,F(xiàn)S-1井的吸水剖面反演結(jié)果與生產(chǎn)測(cè)井工具(PLT)測(cè)量結(jié)果的對(duì)比如圖8所示。其中0~28 m井段為不滲透段(套管段),吸水量為0。圖8表明53~93 m井段為相對(duì)高滲透井段,吸水量最高,其他井段的吸水量相差較小。從圖8中可見(jiàn),各井段吸水量的反演值與測(cè)量值高度吻合,平均相對(duì)誤差約為13%。其中40.5~53,73~93,93~105.5 m井段吸水量的反演值與測(cè)量值的相對(duì)誤差小于2%,證實(shí)了利用此方法反演水平井吸水剖面的可靠性。井筒末端105.5~118 m井段的反演結(jié)果誤差較大,這可能是井筒末端儲(chǔ)層中存在的“橫流”造成的。流體注入到儲(chǔ)層中后,在儲(chǔ)層中發(fā)生橫向流動(dòng),影響了儲(chǔ)層與井筒之間的熱量傳遞,最終導(dǎo)致反演與PLT測(cè)量的結(jié)果存在一定差距。
注入剖面的反演結(jié)果可以反映各井段注入量的相對(duì)貢獻(xiàn)率,指導(dǎo)現(xiàn)場(chǎng)工程師制定更加精細(xì)化的注液方案,并且可根據(jù)各層段相對(duì)注入量的變化,動(dòng)態(tài)調(diào)整注液方案,防止發(fā)生指進(jìn),以達(dá)到最優(yōu)的注入效果。從圖8中可以看出,53~93 m井段的相對(duì)吸水量達(dá)到了54%,相對(duì)滲透率較高,而其他井段的相對(duì)吸水量不超過(guò)15%,在注液過(guò)程中發(fā)生了一定程度的指進(jìn)現(xiàn)象。此時(shí),需要快速制定相應(yīng)的調(diào)剖方案,以實(shí)現(xiàn)最佳注液效果。
3.3 產(chǎn)液剖面反演
將FS-2井的相關(guān)參數(shù)及溫度測(cè)量數(shù)據(jù)代入到反演目標(biāo)函數(shù)(式(19))中進(jìn)行反演計(jì)算,其中反演誤差ε設(shè)置為10-4。當(dāng)反演誤差函數(shù)滿足預(yù)設(shè)精度后,F(xiàn)S-2井的產(chǎn)液剖面反演結(jié)果如圖9所示。其中,191~292 m井段為套管段,產(chǎn)液量為0。反演結(jié)果中452~572 m井段為對(duì)產(chǎn)量貢獻(xiàn)率最高的層段,332~412 m井段次之,而遠(yuǎn)離水平井跟部的632~675 m井段的貢獻(xiàn)率接近為0;這與井筒測(cè)量溫度的變化趨勢(shì)基本一致。這說(shuō)明此SAGD產(chǎn)液井井筒中的溫度主要由儲(chǔ)層液體攜帶至井筒的熱量決定。反演過(guò)程中獲得的反演溫度和熱電偶測(cè)量的井筒溫度對(duì)比如圖10所示。從圖10中可以看出,反演溫度和測(cè)量溫度吻合度較高,平均絕對(duì)誤差小于0.2 ℃,證實(shí)了反演過(guò)程的準(zhǔn)確性。
利用FS-2井口測(cè)量的總產(chǎn)量驗(yàn)證反演結(jié)果,F(xiàn)S-2-1、FS-2-2、FS-2-3、FS-2-4、FS-2-5組總產(chǎn)液量的測(cè)量值分別為87、85、70、103、121 m3,對(duì)應(yīng)的反演值分別為86.99、85.01、69.98、120、109.91 m3??偖a(chǎn)液量的反演值與測(cè)量值吻合度較高,平均相對(duì)誤差為5.13%。其中FS-2-1組、FS-2-2組以及FS-2-3組的相對(duì)誤差小于0.1%,證明該方法可靠。對(duì)比FS-2井各時(shí)間段的產(chǎn)液量(圖9),若在注采施工參數(shù)不變的情況下,此結(jié)果可定性評(píng)價(jià)FS-2井各井段滲透率的動(dòng)態(tài)變化,并用以診斷井筒中可能發(fā)生儲(chǔ)層傷害的井段,為制定增產(chǎn)方案提供依據(jù)。
熱電偶測(cè)溫的缺點(diǎn)在于測(cè)溫點(diǎn)的間距較大,無(wú)法捕捉更為精細(xì)的溫度剖面數(shù)據(jù),使反演的流動(dòng)剖面精細(xì)化程度較低。采用三次插值算法擬合FS-2井熱電偶測(cè)溫結(jié)果并反演流動(dòng)剖面。為方便比較,將總產(chǎn)液量設(shè)為1,得到沿井筒連續(xù)分布的歸一化產(chǎn)液剖面,如圖11所示。從橫坐標(biāo)向上引垂線,即可得到井筒中任一點(diǎn)產(chǎn)液剖面隨時(shí)間的變化。假設(shè)產(chǎn)液過(guò)程中流體流動(dòng)符合達(dá)西滲流規(guī)律,在注采條件不變時(shí),滲透率與產(chǎn)液量成正比??衫脷w一化產(chǎn)液剖面半定量分析滲透率變化。利用達(dá)西滲流公式和產(chǎn)液剖面分別求得井筒中300、400、550與650 m處的歸一化滲透率,如圖12所示。300和650 m處的滲透率隨時(shí)間呈下降趨勢(shì),由此可判斷這兩處地層有發(fā)生儲(chǔ)層傷害的趨勢(shì),應(yīng)及時(shí)制定相應(yīng)的治理措施。
反演結(jié)果精度及連續(xù)性取決于溫度測(cè)量結(jié)果的精度及連續(xù)性。隨著溫度探測(cè)器的快速發(fā)展,未來(lái)可獲取時(shí)間間隔更短、分辨率更高的井筒溫度數(shù)據(jù)(如DTS光纖測(cè)溫所得數(shù)據(jù)),可利用本方法對(duì)地層流動(dòng)剖面及滲透率開(kāi)展更精細(xì)的分析,為現(xiàn)場(chǎng)提供精細(xì)化的配產(chǎn)配注方案。
4 結(jié) 論
(1)建立的反演方法不需要求解復(fù)雜的耦合模型,計(jì)算效率高;不僅可以定量監(jiān)測(cè)流動(dòng)剖面的動(dòng)態(tài)變化,評(píng)價(jià)各層段的貢獻(xiàn)率,還可半定量刻畫水平井各層段相對(duì)滲透率的演化規(guī)律,指導(dǎo)現(xiàn)場(chǎng)制定更加精細(xì)化的配產(chǎn)、配注及增產(chǎn)方案。
(2)Adam優(yōu)化算法同時(shí)利用了梯度的一階矩估計(jì)和二階矩估計(jì),自適應(yīng)調(diào)整反演參數(shù)的步長(zhǎng),具有較高的初速度??筛咝Х囱菟骄鲃?dòng)剖面。
(3)阿曼Safah油田和新疆風(fēng)城油田的兩口水平井的流動(dòng)剖面反演結(jié)果與現(xiàn)場(chǎng)測(cè)量數(shù)據(jù)之間吻合度較高。其中吸水剖面反演結(jié)果與PLT測(cè)量結(jié)果之間的平均相對(duì)誤差為13%,產(chǎn)液剖面反演的日產(chǎn)液量與實(shí)測(cè)值的平均相對(duì)誤差為5.13%,證實(shí)此方法可靠。
參考文獻(xiàn):
[1]侯連華,王京紅,信荃麟,等.吸水剖面測(cè)井資料在油田開(kāi)發(fā)中的應(yīng)用[J].石油大學(xué)學(xué)報(bào)(自然科學(xué)版),1997,21(2):20-22,26,116.
HOU Lianhua, WANG Jinghong, XIN Quanlin, et al. Application of water absorption profile logging data in oilfield development[J]Journal of the University of Petroleum,China(Edition of Natural Science),1997,21(2):20-22,26,116.
[2]隋義勇,李治平,王繼強(qiáng),等.用隱式非線性方法預(yù)測(cè)注水井吸水剖面[J].中國(guó)石油大學(xué)學(xué)報(bào)(自然科學(xué)版),2010,34(6):95-98.
SUI Yiyong, LI Zhiping, WANG Jiqiang, et al. Prediction of water injection profile using implicit nonlinear method[J]. Journal of China University of Petroleum (Edition of Natural Science), 2010,34(6):95-98.
[3]王海靜,薛世峰,高存法,等.非均質(zhì)油藏水平井射孔調(diào)剖方法[J].中國(guó)石油大學(xué)學(xué)報(bào)(自然科學(xué)版),2012,36(3):135-139,150.
WANG Haijing, XUE Shifeng, GAO Cunfa, et al. Profile control method for perforated horizontal wells in heterogeneous reservoirs[J]. Journal of China University of Petroleum (Edition of Natural Science), 2012,36(3):135-139,150.
[4]HASHISH R G, ZEIDOUNI M. Injection profiling in horizontal wells using temperature warmback analysis[J]. Computational Geosciences, 2021,25(1):215-232.
[5]OUYANG L B. Production and injection profiling-challenges and new opportunities[R]. SPWLA-2005-MM, 2005.
[6]KAESHKOV I S, KREMENETSKIY M I, BUYANOV A V. Horizontal well monitoring based on temperature profile measuring with distributed temperature sensors (DST)[R]. SPE 171236-MS, 2014.
[7]OUYANG L B, BELANGER D L. Flow profiling by distributed temperature sensor (DTS) system-expectation and reality[J]. SPE? Production & Operations, 2006,21(2):269-281.
[8]HASHISH R G, ZEIDOUNI M. Analytical approach for injection profiling through warm-back analysis in multilayer reservoirs[J]. Journal of Petroleum Science and Engineering, 2019,182:106274.
[9]HASHISH R G, ZEIDOUNI M. Warmback analysis to determine fracture geometry of a single-stage hydraulic fracturing stimulation[J]. International Journal of Thermal Sciences, 2020,155:106423.
[10]PIMENOV V, BROWN G A, TERTYCHNYI V V, et al. Injectivity profiling in horizontal wells via distributed temperature monitoring[R]. SPE 97023-MS, 2005.
[11]ZHANG S, ZHU D. Efficient flow rate profiling for multiphase flow in horizontal wells using downhole temperature measurement[R]. IPTC 19138-MS, 2019.
[12]YOSHIOKA K, ZHU D, HILL A D, et al. Detection of water or gas entries in horizontal wells from temperature profiles[R]. SPE 100209-MS, 2006.
[13]YOSHIOKA K, ZHU D, HILL A D. A new inversion method to interpret flow profiles from distributed temperature and pressure measurements in horizontal wells[R]. SPE 109749-MS, 2007.
[14]YOSHIOKA K. Detection of water or gas entry into horizontal wells by using permanent downhole monitoring systems[D]. Texas: Texas A&M University, 2007.
[15]SUI W, ZHU D, HILL A D, et al. Determining multilayer formation properties from transient temperature and pressure measurements[R]. SPE 116270-MS, 2008.
[16]朱世琰.基于分布式光纖溫度測(cè)試的水平井產(chǎn)出剖面解釋理論研究[D].成都:西南石油大學(xué),2016.
ZHU Shiyan. Horizontal well production profile interpretation based on distributed temperature sensing[D]. Chengdu: Southwest Petroleum University, 2016.
[17]蔡珺君.水平井井筒溫度預(yù)測(cè)及解釋模型研究[D].成都:西南石油大學(xué),2016.
CAI Junjun. Research on temperature prediction and interpretation model of horizontal well[D]. Chengdu: Southwest Petroleum University, 2016.
[18]CAI J, DUAN Y. Study on temperature distribution along wellbore of fracturing horizontal wells in oil reservoir[J]. Petroleum, 2015,1(4):358-365.
[19]LI Z, ZHU D. Predicting flow profile of horizontal well by downhole pressure and distributed-temperature data for waterdrive reservoir[J]. SPE Production & Operations, 2010,25(3):296-304.
[20]LI Z. Interpreting horizontal well flow profiles and optimizing well performance by downhole temperature and pressure data[D]. Texas: Texas A & M University, 2010.
[21]LUO H, LI H, LI Y, et al. Investigation of temperature behavior for multi-fractured horizontal well in low-permeability gas reservoir[J]. International Journal of Heat and Mass Transfer, 2018,127:375-395.
[22]LUO H, LI H, LU Y, et al. Inversion of distributed temperature measurements to interpret the flow profile for a multistage fractured horizontal well in low-permeability gas reservoir[J]. Applied Mathematical Modelling, 2020,77:360-377.
[23]羅紅文,李海濤,蔣貝貝,等.基于DTS數(shù)據(jù)反演的低滲氣藏壓裂水平井產(chǎn)出剖面解釋新方法[J].天然氣地球科學(xué),2019,30(11):1639-1645.
LUO Hongwen, LI Haitao, JIANG Beibei, et al. A novel method to interpret production profiles of fractured horizontal well in low-permeability gas reservoir by inversing DTS data[J]. Natural Gas Geoscience,2019,30(11):1639-1645.
[24]LUO H, JIANG B, LI H, et al. Flow rate profile interpretation for a two-phase flow in multistage fractured horizontal wells by inversion of DTS data[J]. ACS Omega, 2020,5(34):21728-21744.
[25]LUO H, LI Y, LI H, et al. Simulated annealing algorithm-based inversion model to interpret flow rate profiles and fracture parameters for horizontal wells in unconventional gas reservoirs[J]. SPE Journal, 2021,26(4):1679-1699.
[26]李海濤,羅紅文,向雨行,等.基于DTS的頁(yè)巖氣水平井人工裂縫識(shí)別與產(chǎn)出剖面解釋方法[J].天然氣工業(yè),2021,41(5):66-75.
LI Haitao, LUO Hongwen, XIANG Yuxing, et al. DTS based hydraulic fracture identification and production profile interpretation method for horizontal shale gas wells[J]. Natural Gas Industry, 2021,41(5):66-75.
[27]羅紅文,李海濤,李穎,等.低滲透氣藏壓裂水平井產(chǎn)出剖面與裂縫參數(shù)反演解釋[J].石油學(xué)報(bào),2021,42(7):936-947.
LUO Hongwen, LI Haitao, LI Ying, et al.Inversion and interpretation of production profile and fracture parameters of fractured horizontal wells in low-permeability gas reservoirs[J]. Acta Petrolei Sinica, 2021,42(7):936-947.
[28]MIAH M I, ELHAJ M A, AHMED S, et al. Modeling of temperature distribution and oil displacement during thermal recovery in porous media: a critical review[J]. Fuel, 2018,226:423-440.
[29]KINGMA D P, BA J. Adam: a method for stochastic optimization[J]. arXiv Preprint arViv, 2015: 1412.6980.
[30]孟繁磊,范秦寅,穆麗紅.基于Adam優(yōu)化的二維自適應(yīng)TFPF地震去噪算法[J].吉林大學(xué)學(xué)報(bào)(信息科學(xué)版),2020,38(1):18-26.
MENG Fanlei, FAN Qinyan, MU Lihong. 2D-ATFPF seismic denoising algorithm based on Adam optimization[J]. Journal of Jilin University (Information Science Edition), 2020,38(1):18-26.
[31]田睿,孟海東.基于Adam優(yōu)化算法的深度神經(jīng)網(wǎng)絡(luò)巖爆預(yù)測(cè)模型[J].礦業(yè)研究與開(kāi)發(fā),2020,40(11):40-46.
TIAN Rui, MENG Haidong. Rockburst prediction model of deep neural network based on Adam optimization algorithm[J].Mining Research and Development,2020,40(11):40-46.
[32]費(fèi)致根,吳志營(yíng),肖艷秋,等.基于Adam優(yōu)化算法的雙目機(jī)器人手眼標(biāo)定方法[J].機(jī)床與液壓,2021,49(11):26-30.
FEI Zhigen, WU Zhiying, XIAO Yanqiu, et al. Adam-based hand-eye calibration method for robot with binocular vision[J].Machine Tool & Hydraulics,2021,49(11):26-30.
[33]盧漢,郭大立,李曙光,等.基于CatBoost-Adam結(jié)合的煤層氣產(chǎn)量通用預(yù)測(cè)與優(yōu)化方法研究[J].科技通報(bào),2021,37(7):1-6,12.
LU Han, GUO Dali, LI Shuguang, et al. Research on the general prediction and optimization method of coalbed methane production based on CatBoost-Adam combination[J]. Science and Technology Bulletin, 2021,37(7):1-6,12.
(編輯 李志芬)
收稿日期:2022-06-15
基金項(xiàng)目:國(guó)家自然科學(xué)基金重大項(xiàng)目(51991362)
第一作者:朱海濤(1996-),男,博士研究生,研究方向?yàn)橛蜌夤こ绦畔⒒c智能化技術(shù)。E-mail:zhuhaitao196@163.com。
通信作者:林伯韜(1983-),男,教授,博士,博士生導(dǎo)師,研究方向?yàn)橛蜌夤こ绦畔⒒c智能化技術(shù)。E-mail:linb_cupb@163.com。