• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>SPE eastern regional meeting (ERM 99) >INFILL DRILLING RECOVERY MODELS FOR CARBONATE RESERVOIRS - A MULTIPLE STATISTICAL, NON-PARAMETRIC REGRESSION, AND NEURAL NETWORK APPROACH
【24h】

INFILL DRILLING RECOVERY MODELS FOR CARBONATE RESERVOIRS - A MULTIPLE STATISTICAL, NON-PARAMETRIC REGRESSION, AND NEURAL NETWORK APPROACH

机译：碳酸盐岩储层的钻井钻探恢复模型-多种统计，非参数回归和神经网络方法

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

We used multiple mathematical techniques to developrnprimary ultimate oil recovery (PRUR), initial waterfloodingrnultimate oil recovery (IWUR), and infill drilling ultimate oilrnrecovery (IDUR) models for carbonate reservoirs in WestrnTexas. These are (1) non-linear regresion, (2) non-parametricrnregression (a statistical approach that constructs the functionalrnrelationship between dependent and independent variables,rnwithout bias towards a particular model), and (3) neural networkrnmodels. Development of the oil recovery forecast models helprnunderstand the relative importance of dominant reservoirrncharacteristics and operational variables, reproduce recoveriesrnfor units included in the database, and forecast recoveries forrnpossible new units in similar geological setting.rnOne of the challenges in this research was to identify therndominant and the optimum number of independent variables.rnThe variables is large including porosity, permeability, waterrnsaturation, depth, area, net thickness, gross thicknees, formationrnvolume factor, pressure, viscosity, API gravity, number of wellsrnin initial waterflooding, number of wells for primary recovery,rnnumber of infill wells over the initial waterflooding, PRUR,rnIWUR, and IDUR. The limited number of field data (43 data)rnpoints) are inexact and often exhibit uncertain relationships. Wernused an intelligence software1 that integrates multivariaternstatistical and neural networks to develop accurate neuralrnnetwork models. Multivariate principal component analysis isrnused to identify the dominant and the optimum number ofrnindependent variables. We compared the results from neuralrnnetwork models with the non-parametric approach. The advantage of the non-parametric regression is easy to use andrnthe disadvantage is retaining a large variance of forecast resultsrnfor a particular data set.rnDevelop a neural network model that is an “accurate”rnrepresentation of data may be and ardous task that requirernexperience with the qualitative effects of the structuralrnparameters of neural network models. The advantage of neuralrnnetwork PRUR, IWUR, and IDUR models are capable ofrnforecasting the oil recovery with less error variance.
机译：我们使用多种数学技术开发了WestrnTexas碳酸盐岩储层的初次最终采油量（PRUR），初次注水最终最终采油量（IWUR）和填充钻探终极采油量（IDUR）模型。这些是（1）非线性递归，（2）非参数回归（一种统计方法，可构建因变量和自变量之间的函数关系，而不会偏向特定模型），以及（3）神经网络模型。石油采收率预测模型的开发有助于了解主要储层特征和操作变量的相对重要性，为数据库中包含的单元重现采收率，以及在相似地质环境中预测可能的新单元的采收率。本研究的挑战之一是要确定主要储层和储层。独立变量的最佳数量。变量很大，包括孔隙度，渗透率，水饱和度，深度，面积，净厚度，总厚度，地层体积因子，压力，粘度，API重力，初始注水井数，初次采油井数，初始注水，PRUR，rnIWUR和IDUR上的填充井数量。有限数量的现场数据（43个数据点）是不精确的，并且经常表现出不确定的关系。 Wernus使用了一种智能软件1，该软件集成了多变量统计和神经网络以开发准确的神经网络模型。使用多变量主成分分析来确定独立变量的主要和最佳数量。我们将神经网络模型的结果与非参数方法进行了比较。非参数回归的优点是易于使用，缺点是保留了特定数据集的预测结果的较大差异。开发一种神经网络模型，即“准确”地表示数据可能是艰巨的任务，需要对数据进行经验性的神经网络模型结构参数的定性影响。神经网络PRUR，IWUR和IDUR模型的优点是能够以较小的误差方差预测油采收率。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号