大庆石油地质与开发2025，Vol.44Issue(5)：62-69,8.DOI:10.19597/J.ISSN.1000-3754.202504056
基于大数据分析的特高含水后期水淹层智能解释方法
Intelligent interpretation method for water flooded reservoirs in late stage of ultra-high water cut based on big data analysis
摘要
Abstract
In late stage of ultra-high water cut,underground fluid properties become increasingly complex,exerting a growing influence on well logging curves.In view of the problems of the traditional interpretation method including low utilization rate of well logging curves,unsatisfactory interpretation coincidence rate for production requirements and low interpretation efficiency,a"knowledge+data"dual driven intelligent logging evaluation technique for water flooded zone is proposed based on big-data analysis.Through techniques such as multi-sources data fusion,abnor-mal curves processing and spontaneous potential(VSP)mudstone baseline construction,a preprocessing method for big data analysis is established.By innovatively integrating the expert experience,well logging professional knowl-edge and data,a multi-dimensional feature characterizing technique is developed based on feature engineering guid-ed by logging mechanisms.Over 10 machine learning algorithms are selected to build a water-flooded reservoir iden-tification model,achieving intelligent evaluation for water-flooded reservoirs.Actual application results indicate that the new method,validated in 23 sealed coring wells,significantly improves the prediction accuracy of reservoir parameters,with an average relative error of 5.76％for porosity and an average absolute error of 7.03％for current water saturation.Compared to previous methods,the prediction accuracy of reservoir parameters has been signifi-cantly improved,with average relative error of porosity decreased by 2 percentage points and current average abso-lute error of water saturation decreased by 1 percentage point.The coincidence rate of thin-layer water flooded grades increases from 70.0％with the traditional interpretation method to 77.1％,while the coincidence rate of thick-layer water flooded grades increases from 75.0％with the traditional interpretation method to 81.4％.The research provides a basis for the perforation and profile adjustment measures,offering strong technical support for remaining oil evaluation,further quality and efficiency improvements and potential tapping.关键词
特高含水后期/水淹层智能解释/大数据分析/机器学习算法/参数预测Key words
late stage of ultra-high watercut/intelligent interpretation of water-flooded reservoir/big data analy-sis/machine learning algorithm/parameter prediction分类
天文与地球科学引用本文复制引用
卢艳,马宏宇,齐云峰,程梦薇,刘宇轩,王雪萍..基于大数据分析的特高含水后期水淹层智能解释方法[J].大庆石油地质与开发,2025,44(5):62-69,8.基金项目
中国石油天然气集团有限公司"十三五"科技开发基金项目"高-特高含水油田改善水驱效果关键技术"(2019B-1209). （2019B-1209）
