中国石油勘探 ›› 2020, Vol. 25 ›› Issue (5): 97-103.DOI: 10.3969/j.issn.1672-7703.2020.05.013

• 工程技术 • 上一篇    下一篇

基于梦想云的测井智能化解释应用研究

赵丽莎1,史永彬2,3,金 玮2,3,李 华1,塔斯肯1   

  1. 1 中国石油勘探开发研究院;2 中国石油集团东方地球物理勘探有限责任公司;3 北京中油瑞飞信息技术有限责任公司
  • 出版日期:2020-09-14 发布日期:2020-09-12
  • 基金资助:
    中国石油天然气股份有限公司投资信息化重点项目“勘探开发一体化协同研究及应用平台(一期)建设”(PetroChina- IT-2017-N104)。

Application research on intelligent logging interpretation based on E&P Dream Cloud

Zhao Lisha1,Shi Yongbin2.3, Jin Wei2.3, Li Hua1, Ta Siken1   

  1. 1 PetroChina Research Institute of Petroleum Exploration & Development; 2 BGP Inc., CNPC; 3 Richfi t Information Technology Co., Ltd.
  • Online:2020-09-14 Published:2020-09-12
  • Supported by:
     

摘要: 迄今为止,地球物理测井仍然是石油工业中用于识别地下岩层物性以及是否含油气性的最重要的技术手 段。利用大数据和人工智能方法,基于梦想云智能数据分析平台开展测井智能化解释,以数据收集和拼接为基础,进 行特征抽取,结合专家知识和经验构建模型,开展机器学习和模型训练,完成模型测试与迭代,并分别对砂岩、泥岩 的判别(阈值为0.4,60口井)和油层、水层判别(阈值为0.5,47口井)进行了智能化解释技术验证。解释结果显示: 砂岩、泥岩判别用时约2~4h,泥岩和砂岩判别准确率分别为94%和90%,综合判别准确率为92%;油层、水层判 别用时约 4~6h,油层和水层判别准确率分别为93%和70%,综合判别准确率为86%,有效提高了多口井测井解释 自动化程度和结果符合率,大数据分析技术应用为传统的石油测井解释技术发展带来了新的动能和良好的发展前景。

 

关键词: 梦想云, 测井, 大数据, 人工智能, 测井解释

Abstract: Geophysical logging is still the most important technical means used to identify the physical properties of subsurface strata and to decide whether they are hydrocarbon-bearing. Using the technologies of big data and artificial intelligence, based on the E&P Dream Cloud intelligent data analysis platform, this paper presents a new approach to intelligent well logging interpretation. Features of log data are extracted on the basis of data collection and splicing. Expert knowledge and experience are combined to build model. Machine learning and model training are carried out, and model testing and iteration are completed. For this study, intelligent logging interpretation technology was verified by confirming sandstone and mudstone differentiation (with a cut-off value of 0.4 based on 60 wells) and oil and water layer differentiation (with a cut-off value of 0.5 based on 47 wells). The interpretation results show that the intelligent system takes about 2-4 hours to distinguish between sandstone and mudstone. The accuracy rate of mudstone and sandstone identification is 94% and 90% respectively with an overall accuracy rate of 92%. It takes about 4-6 hours to identify oil and water layers. The accuracy rate of oil and water layer identification is 93% and 70% respectively with an overall accuracy rate of 86%. This technology can effectively increase the automation degree and the coincidence rate of multi-well logging interpretation. The application of big data analysis technology provides a new motive force and good prospects for the development of well logging interpretation technology.

Key words: E&, P Dream Cloud, well logging, big data, artificial intelligence, well log interpretation

中图分类号: