摘要
研究表明,智能识别方法能够有效提高早期溢流监测的准确性,但由于溢流是钻井过程中的小概率事件,现场可获取的溢流样本数据十分有限,限制了智能识别方法的应用。针对该问题,在分析常用溢流监测参数与钻井设计参数、地质相关参数以及钻井工况之间关系的基础上,建立了钻进、起钻、下钻三种工况下常用溢流监测参数的数值模拟模型,为溢流智能识别方法的应用提供数据基础。利用现场实测数据对所建模型进行了验证实验。结果表明:在给定的钻井条件下,仿真得到的数据与现场实测数据间的相似度达到83.85%,具有较高的吻合度;溢流智能识别模型识别准确率较缺乏训练样本的专家经验模型提高了23.1%,识别准确率得到了显著提升。
The research shows that the intelligent identification method can effectively improve the accuracy of early overflow monitoring.However,because the overflow is a small probability event in drilling process,the overflow sample data available in the field is very limited,which limits the application of intelligent identification method.To solve this problem,based on the analysis of the relationship between common overflow monitoring parameters and drilling design parameters,geological related parameters and drilling conditions,the numerical simulation models of commonly used overflow monitoring parameters under three conditions of drilling,come out of the hole and go in the hole are established to provide data basis for the application of intelligent overflow identification method.The models are verified by the measured data in the field.Under the given drilling conditions,the similarity between the simulated data and the field measured data is up to 83.85%,with a high degree of coincidence.The identification accuracy of overflow intelligent identification model is 23.1%higher than that of the expert empirical model without training samples,and the identification accuracy is significantly improved.
作者
张博
胡旭光
刘贵义
李宜君
孙伟峰
戴永寿
Zhang Bo;Hu Xuguang;Liu Guiyi;Li Yijun;Sun Weifeng;Dai Yongshou(College Of Information and Control Engineering,China University Of Petroleum(East China),Huangdao District,Qingdao City,Shandong Province,266580,China;不详)
出处
《录井工程》
2019年第4期44-50,147,共8页
Mud Logging Engineering
基金
中国石油天然气集团公司重大科技项目“油气井井喷预防与控制技术研究及应用”(编号:2016D-4601)