王雪飞,高科超,张兴华,罗鹏,杜连龙,刘境玄. 试井数据监督式机器学习方法预测变质岩潜山储层产量[J]. 石油钻采工艺,2024,46(4):443-454. DOI: 10.13639/j.odpt.202409051
引用本文: 王雪飞,高科超,张兴华,罗鹏,杜连龙,刘境玄. 试井数据监督式机器学习方法预测变质岩潜山储层产量[J]. 石油钻采工艺,2024,46(4):443-454. DOI: 10.13639/j.odpt.202409051
WANG Xuefei, GAO Kechao, ZHANG Xinghua, LUO Peng, DU Lianlong, LIU Jingxuan. A supervised machine learning method based on well test data to predict reservoir production in buried-hill metamorphic rock[J]. Oil Drilling & Production Technology, 2024, 46(4): 443-454. DOI: 10.13639/j.odpt.202409051
Citation: WANG Xuefei, GAO Kechao, ZHANG Xinghua, LUO Peng, DU Lianlong, LIU Jingxuan. A supervised machine learning method based on well test data to predict reservoir production in buried-hill metamorphic rock[J]. Oil Drilling & Production Technology, 2024, 46(4): 443-454. DOI: 10.13639/j.odpt.202409051

试井数据监督式机器学习方法预测变质岩潜山储层产量

A supervised machine learning method based on well test data to predict reservoir production in buried-hill metamorphic rock

  • 摘要: (目的意义)建立储层产量预测模型,实现产量定量评价,解决现有技术定性预测无法数字量化产量难题,提高产量预测精度。(方法过程)将初步试井解释渗透率与求产第一油嘴产量两个数据作为研究对象,通过机器学习、数据建模,对初步试井解释渗透率与求产第一油嘴产量之间的客观规律进行数学表达,建立函数关系,形成产量定量评价公式。依托变质岩潜山储层 5 口井有效测试数据进行自助法验证,同时在 2 口井测试作业现场进行验证,评价数据模型预测产量的准确度。(结果现象)应用自助法分两次抽取( 3,3 )个数据输入数据模型公式,通过公式预测的单口井产量与实际单井测试第一油嘴求产产量的相对误差为1%~ 10%,证明了该数据模型的有效性和稳定性。同时,在现场两口井的产量预测中使用该公式,相对误差分别为8 %与 16 %,再次证明了产量评价模型对新数据具有适应性,数据模型可靠有效。(结论建议)为变质岩潜山储层产量定量预测提供了一种可选择的方法。

     

    Abstract: The vertical and radial heterogeneity of multi-zone dispersed reservoirs is strong, which brings difficulties to reservoir evaluation. Well test dynamic characteristics machine learning technology, improve the evaluation level. This article analyzes the double logarithmic curves of 10 wells/layers in the complex reservoir of Bohai A structure's ancient, buried hill, and forms the characteristic curve of this complex reservoir. At the same time, through the fusion data analysis technology, the typical feature template of exploration and well testing in the complex reservoir of Bohai A structure's ancient, buried hill is formed by analyzing the well testing parameters of the complex reservoir. The study suggests that the ancient, buried hill test wells in Bohai A structure belong to the same reservoir, and it also proves that the continuity of the reservoir is good. At the same time, this article successfully achieved node-based production capacity prediction through regional well testing results and provided a production capacity prediction formula. The reverse validation of the prediction model achieved a compliance rate of 75%. Through this study, a dynamic interpretation template for the ancient buried hill test of Bohai A structure was provided, which achieved the identification of the dynamic change trend of the oil reservoir. The regional well testing method studied in this article is derived from traditional regional well testing techniques and effectively combines data analysis techniques, which can broaden the application scenarios of well testing.

     

/

返回文章
返回