Selection of decline curve analysis model using Akaike information criterion for unconventional reservoirs
Selection of decline curve analysis model using Akaike information criterion for unconventional reservoirs
复制标题
DOI：
10.1016/j.petrol.2019.106327
复制
发表时间：
2019-11
影响因子：
--
通讯作者：
A. Shabib-Asl;Tatyana Plaksina
中科院分区：
文献类型：
--
作者：
A. Shabib-Asl;Tatyana Plaksina
来自非常规资源（特别是致密油藏和页岩油藏）的石油生产已成为北美供应的重要组成部分，这促使运营公司寻找更准确的储量评估和产量预测方法。通常，运营商采用递减曲线分析 (DCA) 来确定常规和非常规油藏中碳氢化合物储量的估计最终采收率 (EUR)。在页岩井中，最佳 DCA 模型适合瞬态流态，其次是边界支配流 (BDF)。因此，在本研究中，我们提出了一种新的系统 DCA 模型选择框架，使用页岩井生产数据拟合优度的定量测量。为此，我们选择了七种不同的、最常用的 DCA 模型来评估 Montney 地层的大量井。为了收集 Montney 井的生产数据，我们使用了 GeoSCOUT 软件包。然后，我们将这些数据输入由 Nelder-Mead 单纯形算法支持的 DCA 模型选择工作流程，并使用计算机编程语言实现。有助于确定数据和模型最佳拟合的定量测量源自信息论 (IT)，称为赤池信息准则 (AIC)。获得的结果表明，我们的 DCA 模型选择框架可以为给定的生产历史识别最合适的 DCA 模型。此外，它还给出了可用于模型排名的特定拟合定量度量（AIC）。结果表明，Arps 的 DCA 模型需要更准确的 b 值来匹配 Montney 页岩井的生产数据，并确定 Logistic 增长分析模型 (LGM) 最适合大多数选定的井，其次是扩展指数递减模型 (EED)、Duong 和幂律指数模型 (PLE)。
Petroleum production from unconventional resources (specifically, tight and shale reservoirs) has become a significant portion of the supply in North America which led operating companies to look for more accurate methods for reserve evaluation and production forecasting. Typically, the operators resort to Decline Curve Analysis (DCA) to determine the Estimated Ultimate Recovery (EUR) of hydrocarbon reserves in both conventional and unconventional reservoirs. In shale wells, the best DCA model fitted to the transient flow regimes followed by Boundary Dominated Flow (BDF). Therefore, in this study, we propose a novel framework for systematic DCA model selection using a quantitative measure of goodness of production data fit from shale wells. For this purpose, we selected seven different, most frequently applied DCA models to evaluate large number of wells from the Montney Formation. To collect production data from the Montney wells, we used the GeoSCOUT package. Then, we fed these data into the DCA model selection workflow powered by the Nelder-Mead simplex algorithm and implemented using a computer programming language. The quantitative measure that helps identify the best fit of the data and the model is derived from the Information Theory (IT) and is known as Akaike Information Criterion (AIC). The obtained results show that our DCA model selection framework can identify the most appropriate DCA model for a given production history. Moreover, it gives a specific quantitative measure (AIC) of the fit that can be used for model ranking. The results show that more accurate values of b are necessary in Arps’ DCA models to match production data from the Montney shale wells, and identify the Logistic Growth Analysis Model (LGM) is the best fit for the majority of the selected wells followed by Extended Exponential Decline Model (EED), Duong, and Power Law Exponential Model (PLE).