基于气测数据的钻井溢流智能实时预警模型  

Intelligent real-time early warning model for drilling overflow based on gas logging data

作　　者：管震 李雪松 王建华 覃吉 胡宏涛 李洪松 谢冬 GUAN Zhen;LI Xuesong;WANG Jianhua;QIN Ji;HU Hongtao;LI Hongsong;XIE Dong(Kunlun Digital Intelligence Technology Co.,Ltd.,CNPC,Beijing 100043,China)

机构地区：[1]昆仑数智科技有限责任公司,北京100043

出　　处：《世界石油工业》2024年第5期83-90,共8页World Petroleum Industry

基　　金：中国石油天然气集团有限公司“钻完井及井下作业智能优化系统研发”(2021DJ7401)。

摘　　要：溢流是井喷的重要先兆,及时发现早期溢流是成功且有效控制井喷的直接途径和关键手段。针对钻井过程中溢流风险难以及时发现的问题,应用随钻实时数据,为实现溢流风险智能预警开展积极探索。地层流体中的气体以破碎气、扩散气、渗滤气的形式侵入井筒,表现出不同的气测数据特征。破碎气的气测曲线表现为“山峰状”或“箱状”,扩散气气测曲线表现为基值抬升,渗滤气气测曲线一般表现为“箱状”。通过深入研究地层流体进入井筒方式及气测数据表现特征,分析总结“高压低渗”“单根气监测分析”溢流预警机理,形成样本学习案例集,建立溢流风险智能预警全连接神经网络模型,进而实现预警自动化、智能化。测试表明预警模型能够适应大量溢流井诱发溢流的地质工程条件,尤其适于非常规油气井的溢流监测,可以动态反映液柱压力与地层压力之间的平衡关系。在溢流风险发生之初及时准确提示溢流风险,指导钻井现场有效防控溢流,实现溢流发现处置到主动防范的转变。基于气测数据的溢流智能预警模型,针对钻井溢流风险实时监控分析,做了积极且富有成效的探索,具有推广应用价值。Overflow is an important precursor of blowout.Timely detection of early overflow is the direct and critical way to successfully and effectively control blowout.In response to the challenge of timely detection of drilling overflow risks,active exploration is conducted to achieve intelligent early warning of overflow risks using real-time data during drilling.Hydrocarbons in the formation fluid invade the wellbore in the form of fractured gas,diffused gas,and percolated gas,exhibiting different characteristics of gas logging data.The gas logging curve of fractured gas appears as"peak-shaped"or"box-shaped",the curve of diffused gas shows a rise in baseline,and the curve of percolated gas generally appears as"box-shaped".Through in-depth study of the ways in which formation fluid enters the wellbore and the characteristics of gas logging data,the early warning mechanisms of overflow risks for"high pressure and low permeability"and"connection gas monitoring analysis"are analyzed and summarized.A sample learning case set is formed,and a fully-connected neural network model for intelligent early warning of overflow risks is established.Furthermore,this can lead to the automation and intelligence of early warning systems.Tests indicate that the early warning model can adapt to a wide range of geological and engineering conditions that induced overflow risks in wells,especially suitable for overflow monitoring in unconventional wells.It can dynamically reflect the balance relationship between hydraulic pressure and formation pressure.At the beginning of overflow risks,accurately alarm the overflow risks in time,and guide the effective prevention and control of overflow on the drilling site.This enables the transition from overflow detection to active prevention.The intelligent early warning model of overflow based on gas logging data has conducted active and effective exploration for real-time monitoring and analysis of drilling overflow risks,demonstrating high promotion and application value.

关 键 词：钻井 溢流 智能预警 气测数据 高压低渗 单根气 全连接神经网络 

分 类 号：TE28[石油与天然气工程—油气井工程]