基于EEMD高斯过程自回归模型的缝洞型油藏开发动态指标预测

详细信息    查看全文 | 推荐本文 |

英文篇名：Prediction of Development Dynamic Indexs for Fractured-Vuggy Carbonate Reservoir Based on EEMD and Gaussian Process Autoregression Model 作者：张冬梅 ; 林子航 ; 康志江 ; 王吉祥 ; 邢路通 英文作者：Zhang Dongmei;Lin Zihang;Kang Zhijiang;Wang Jixiang;Xing Lutong;School of Computer Science, China University of Geosciences(Wuhan);Petroleum Exploration and Production Research Institute of SINOPEC (PEPRIS); 关键词：缝洞型油藏 ; 开发动态指标预测 ; 集合经验模态分解 ; 信息熵 ; 高斯过程自回归模型 英文关键词：fractured-vuggy carbonate reservoir;;prediction of development dynamic index;;ensemble empirical mode decomposition(EEMD);;information entropy;;Gaussian process autoregression model 中文刊名：DZKQ 英文刊名：Geological Science and Technology Information 机构：中国地质大学(武汉)计算机学院;中国石化石油勘探开发研究院; 出版日期：2019-05-15 出版单位：地质科技情报 年：2019 期：v.38;No.186 基金：国家科技重大专项“缝洞型油藏提高采收率技术”(2016ZX05014-003) 语种：中文; 页：DZKQ201903028 页数：8 CN：03 ISSN：42-1240/P 分类号：262-269

摘要

缝洞型油藏储集空间类型多样,大缝大洞的存在使得见水特征复杂多样,同时受各类工程、地质因素影响,生产数据非线性、非稳态,动态指标实时预测难度大。对此提出了一种结合集合经验模态分解(EEMD)和信息熵的高斯过程自回归模型的开发动态指标预测方法:①利用EEMD方法将生产数据分解成若干个平稳的本征模态函数(IMF)分量;②采用信息熵计算由于工作制度频繁调整而引起的数据波动程度;③利用分解的低频分量提取拟稳态数据段,对方差贡献度较大的各IMF分量建立高斯过程自回归模型;④叠加各分量计算结果作为预测值。仿真实验表明这种新算法能够有效应用于缝洞型油藏开发动态指标预测,可以预测生产井各项生产指标的变化趋势,为后期生产开发方案调整提供依据,指导油田的整体开发。
        There are various types of reservoir space for fractured-vuggy carbonate reservoir. Large holes and cracks make the water′s characteristics diverse and complicated, which are affected by various engineering and geological factors. As a matter of fact, the nonlinear and non-stationary random signals of the production data series, make the real-time prediction of dynamic indexs difficult. This paper presents a method of Gaussian process autoregression model based on ensemble empirical mode decomposition(EEMD) and information entropy. Firstly, the EEMD method is used to decompose the production data into several stationary intrinsic mode function(IMF) components. Secondly, the information entropy is used to measure the fluctuation degree of data caused by the frequent adjustments of working system. Then, the model is established for every IMF component with superior variance contribution. The model bases on quasi-steady data segment extracted by low-frequency components. Finally, the predicted value is the superposition of each component result. The simulation results show that the new algorithm can be applied to fractured-vuggy carbonate reservoir to predict the variation tendency of every development dynamic target of production wells, which provide basis for adjustment of development plan in later stage production, and guide overall development of oilfield.

引文

[1] 康玉柱.中国古生代碳酸盐岩古岩溶储集特征与油气分布[J].天然气工业,2008,28(6):1-12.
    [2] 李溢龙,吴锋,李晓平,等.塔河缝洞型油藏产量变化预测模型研究[J].石油地质与工程,2016,30(6):26-29.
    [3] Arps J J.Analysis of decline curves[C]//Anon.Petroleum transactions.[S.l.]:[s.n.],1945:228-247.
    [4] Wu Y S,Di Y,Kang Z,et al.A multiple-continuum model for simulating single-phase and multiphase flow in naturally fractured vuggy reservoirs[J].Journal of Petroleum Science & Engineering,2011,78(1):13-22.
    [5] Wu Z,Huang N E.A study of the characteristics of white noise using the empirical mode decomposition method[J].Proceedings Mathematical Physical & Engineering Sciences,2004,460:1597-1611.
    [6] Shannon C E.A mathematical theory of communication[J].Bell Labs Technical Journal,1948,27(4):379-423.
    [7] Borda M.Fundamentals in information theory and coding[J].Vnitrní Lékarství,2011,19(9):883-887.
    [8] Wu Z,Huang N E.Ensemble empirical mode decomposition:A noise-assisted data analysis method[J].Advances in Adaptive Data Analysis,2011,1(1):1-41.
    [9] 陈昌楠.CPI波动结构分解与短期预测研究:基于EEMD模型[J].中国市场,2014(12):47-50.
    [10] Pincus S.Approximate entropy (ApEn) as a complexity measure[J].Chaos,1995,5(1):110-117.
    [11] Seeger M A.Gaussian process for machine learning[J].International Journal of Neural Systems,2004,14(2):69-106.
    [12] Kuss M,Rasmussen C E.Assessing approximate inference for Binary Gaussian process classification[J].Journal of Machine Learning Research,2005,6:1679-1704.