摘要
致密油藏压裂甜点的准确预测是合理部署井位和压裂改造成功的关键。玛湖致密砾岩油藏复杂的地质特征和强非均质性,导致其压裂甜点的预测较为困难。针对目前缺乏有效预测玛湖砾岩油藏压裂甜点方法的问题和提高水平井压裂改造效果的迫切需求,在压裂改造效果主控因素分析基础上,以储层改造体积(stimulated reservoir volume,SRV)为预测指标,首先优选已有基于可压性指数的压裂甜点评价模型,同时建立了基于机器学习的致密砾岩油藏压裂甜点预测模型,最终形成了适用于玛湖致密油藏压裂甜点的预测方法。结果表明:在基于可压性指数的压裂甜点预测模型中,文献[8,10-11]所建立的模型与实际监测结果具有较高吻合度;在基于机器学习算法的压裂甜点预测模型中,随机森林、GRBT(gradient boosting for regression)、Bagging模型表现出较好的性能;虽然当前数据下基于可压性计算的压裂甜点模型的性能较好,但是随着现场数据的更新与准确度的提高,基于机器学习的压裂甜点模型预测精度将不断改善。研究结果对于玛湖致密砾岩油藏压裂甜点和综合甜点评价、井位部署、压裂改造设计具有重要的指导意义。
Accurate prediction of fracturing sweet spots in tight reservoirs is the key to rational well placement and successful stimulation.It is difficult to predict the fracturing sweet spot of Mahu tight conglomerate reservoir due to its complex geological characteristics and strong heterogeneity.In view of the lack of effective methods to predict the fracturing sweet spot of the Mahu conglomerate reservoir and the urgent need to improve the fracturing effect of horizontal wells,stimulated reservoir volume(SRV)was used as a prediction index based on the analysis of main controlling factors of fracturing effects.Firstly,the existing evaluation models of fracturing sweet spot based on the fracability were optimized.The prediction models of fracturing sweet spot based on machine learning were established for tight conglomerate reservoir.Finally,the prediction method for fracturing sweet spot was formed for Mahu tight reservoir.The results show that the models established by ref.[8,10-11]have high precision in the prediction models of fracturing sweet spots based on the fracability.Among fracture sweet spot prediction models based on machine learning algorithm,random forest,gradient boosting for regression(GRBT)and Bagging models show good performance.Although the performance of the fracturing sweet spot model based on the fracability calculation is better,the prediction accuracy of the fracturing sweet spot model based on machine learning will continue to improve as the field data is updated and the accuracy improves.The research results have important guiding significance for evaluation of fracturing sweet spots and comprehensive sweet spots,well placement and fracturing stimulation design of Mahu tight conglomerate reservoir.
作者
杨琨
罗山贵
花凌旭
唐慧莹
孙正龙
YANG Kun;LUO Shan-gui;HUA Ling-xu;TANG Hui-ying;SUN Zheng-long(Xinjiang Oilfield,PetroChina,Karamay 834000,China;State Key Laboratory of Oil and Gas Reservoir Geology and Exploitation,Southwest Petroleum Lniversity,Chengdu 610500,China)
出处
《科学技术与工程》
北大核心
2022年第32期14174-14183,共10页
Science Technology and Engineering
基金
国家自然科学基金(51904257)。
关键词
致密油藏
主控因素
压裂甜点
可压性
机器学习
tight reservoir
main control factors
fracturing sweet spot
fracability
machine learning
