機理-數(shù)據融合的造斜率智能預測方法

打開文本圖片集

中圖分類號：TE243文獻標識碼：ADOI：10.12473/CPM.202404089

Bai Jiashuai，Zhong Yinming，Wang Liwei，etal.Mechanism-data fused intelligent prediction of build-up rate[J].China Petroleum Machinery，2025，53（5）：1-9.

Mechanism-Data Fused Intelligent Prediction of Build-Up Rate

Bai Jiashuai1Zhong Yinming1Wang LiweiLi Zhen2Song Xianzhi2Liu Zihao2Zhu Zhaopeng3 （1.OiProductionTechnology ResearchInstitute（Supervision Company），PetroChina XinjiangOileldompany；2.Collgeof PetroleumEngineering，China UniversityofPetroleu（Beijing）；3.ColegeofMechanicalandTransportationEngineering，Chna U niversity of Petroleum （Beijing））

Abstract：The accurate prediction ofbuild-up rate is fundamental for wellbore trajectory control，and it directlyaffects thedrilling effciencyof directional wells.However，due to the complexityof downhole mechanical behaviors，traditional prediction methods cannot accurately predictthe build-uprate for their limitations.This paper presents a mechanical-intellgent model fused build-up rate prediction method.Specifically，the mechanical model isused tocalculate the bit side force，bit rotation angleand ultimate build-uprate as the main control factors，and the automated machine learning framework is combined with other parameters for purpose of fiting prediction，therebyavoiding theempirical coeficient inversion intraditional method.In this way，theadvantagesof accurate macroscopic pattern description of the mechanical model and strong nonlinear fiting abilityof the inteligent modelare fully exerted.The data from 14 wels inthe Mahu block of Xinjiang Oilfield were used for training and test.The results show that，with mechanical parameters fused，the model achieves the maximum error，root mean square error and average absolute error of build-up rate reduced by 17% ， 12% and 8% ，respectively，with the root mean square error and average absolute error less than 1.00° per 30m ，indicating that this method can effectively improve theaccuracyof build-uprate prediction，and is more effectively performed especially in hole sections with sharp changes in build-up rate.This study provides a new idea for the acurate prediction of build-up rate， and also a technical support for the precise control of wellbore trajectory.

Keywords：intellgent driling；automated machine learning；build-up rate prediction；mechanism-data fused; mechanical parameter

0引言

造斜率的精確預測是實現(xiàn)井眼軌跡高質量調控的基礎，但由于造斜率的影響因素極為復雜，至今尚未完全厘清底部鉆具組合（BHA）在復雜地層中的造斜機理，所以精確預測仍具挑戰(zhàn)性。（剩余13253字）