摘要
运用色谱指纹技术对原油进行产能预测越来越得到油田等相关部门的重视.如何有效地建立原油色谱指纹与产能贡献的模型一直以来是基于色谱指纹的产能预测面临的问题.先前研究者提出的产能预测方法都是基于单一回归模型方法,但是单一的回归方法性能有限,近年来有关回归集成的理论的研究受到了极大的重视,是目前回归领域研究的前沿方向.另外广义可加模型是一种广泛使用的非线性回归模型,不仅适合于建立线性数据的模型,也适合建立非线性数据的模型.因此本文提出对广义可加模型进行集成的方法,使用集成广义可加模型来建立色谱指纹与产能贡献关系的模型.对原油产能预测的实验结果表明,本文提出的方法相对于以前的预测方法的性能有较大的提升,具有较好的预测效果,为解决石油产能预测问题提供了有效的途径.
It is more and more important to oilfield Using chromatographic fingerprint estimation single- zone productivity contribution. However how to effectively build a regression model is still an unsolved problem. Previous researchers have proposed many methods to build the regression model. However all these methods are single regression model. Recently, ensemble regression Model has been obtained a great attention. And generalized additive model is a widely used non-linear regression model. In this pa- per, a method integrating generalized additive model is proposed. The proposed method establishes rela- tionship between the chromatographic fingerprint and the single-zone productivity contribution. Experi- mental results show that the proposed method outperforms other methods, and provide an effective way to estimate the single-zone productivity contribution.
出处
《四川大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第1期79-84,共6页
Journal of Sichuan University(Natural Science Edition)
基金
"油气藏地质及开发工程"国家重点实验室(成都理工大学)资助项目(PLC200902)
关键词
合采井
产能贡献
色谱指纹
广义可加模型
集成回归
commingled oil well, productivity contribution, chromatography fingerprinting, generalizedadditive model, ensemble regression model
