压力与规则双约束下的溢流智能预警模型  

Intelligent Warning Model for Overflow Under Constraints of Pressure and Rules

作　　者：段世明 宋先知[1] 姚学喆 祝兆鹏 许争鸣 Duan Shiming;Song Xianzhi;Yao Xuezhe;Zhu Zhaopeng;Xu Zhengming(State Key Laboratory of Petroleum Resources and Prospecting,China University of Petroleum(Beijing);School of Energy Resources,China University of Geosciences(Beijing))

机构地区：[1]中国石油大学(北京)油气资源与探测国家重点实验室 [2]中国地质大学(北京)能源学院

出　　处：《石油机械》2025年第1期1-9,共9页China Petroleum Machinery

基　　金：国家重点研发计划项目“复杂油气智能钻井理论与方法”(2019YFA0708300);国家自然科学基金杰出青年科学基金项目“油气井流体力学与工程”(52125401);国家自然科学基金青年科学基金项目“深井气侵自动压井的井口回压传播及多相流动特性研究”(52204020);师资博士后科研启动基金项目“深井气侵自动压井井底压力智能预测和控制方法”(2462022SZBHO02)。

摘　　要：随着油气勘探开发向复杂油气资源领域拓展,地层压力层系变化更为复杂。溢流作为钻井高危风险之一,对其进行及时准确预警愈发重要。分析了正常工况与溢流下的表征参数,以逻辑规则构建工况识别模型,实现钻进、起下钻等8种工况的实时分析,从而减少钻井工况对溢流分析的影响。通过特征工程构建数据库,分析数据有效性,利用神经网络形成溢流智能预警模型。利用小波分析诊断溢流波动,结合传统溢流诊断规则,实现压力信号与规则双约束下的溢流智能预警方法,进一步提高模型准确率与泛化能力。通过南海溢流数据与四川盆地实时钻井数据对模型验证,结果表明,受约束后溢流预警模型相较于传统诊断方法可有效提前警报,相较于单一智能模型可有效降低虚警,模型最终准确率为95.28%。该模型有望辅助实现安全高效钻井作业,为智能化钻井提供基础。As oil and gas exploration and development extend towards complex reservoirs,the changes in formation pressure are more complicated.Overflow is one of the high risks in drilling,and conducting timely and accurate warning of it is becoming increasingly important.In the paper,first,the characterization parameters under normal working conditions and overflow were analyzed,and a working condition identification model was built based on logic rules,achieving real-time analysis of 8 working conditions such as drilling and tripping,thereby reducing the influence of drilling conditions on overflow analysis.Second,a database was constructed based on feature engineering to analyze data validity,and the neural network technique was used to build an intelligent warning model of overflow.Third,the wavelet analysis was used to diagnose overflow fluctuations,and combined with traditional overflow diagnosis rules,an intelligent warning method for overflow under constraints of pressure signals and rules was proposed to further improve the accuracy and generalization ability of the model.Finally,the model was verified using the overflow data of the South China Sea and the real-time drilling data of the Sichuan Basin.The results show that the constrained overflow warning model can effectively provide early warnings compared to traditional methods,and can effectively reduce false alarms compared to a single intelligent model.The final accuracy of the model is 95.28%.This model is expected to assist in achieving safe and efficient drilling operations,and provides a foundation for intelligent drilling.

关 键 词：钻井风险 溢流预警 智能模型 压力信号 规则约束 工况识别 

分 类 号：TE21[石油与天然气工程—油气井工程]