应用人工神经网络模型估算不同气体组分和不同储集层原油的最小混溶压力

xinmeng

应用人工神经网络模型估算不同气体组分和不同储集层原油的最小混溶压力

Advances in Geo-Energy Research, 2019, 3(1): 52-66

原文网址:  http://www.astp-agr.com/index.php/Index/Index/detail?id=96

       最小混溶压力（MMP）是监测储层流体与注入气体混溶度的关键变量。可以采用实验和非实验的方法来估算MMP值。已有的混相关联式可以尝试用来预测特定类型气体的最小混溶压力。本文将人工神经网络（ANN）模型应用于来自世界各地251份数据记录的数据组中，提出了一种新方法来估计各种注入气体和原油成分的天然气-原油MMP值。这种方法适用于注入气体成分非常复杂的固存项目。实验模型与油藏温度、原油中挥发性组分（C1和N2）和中间组分（C2、C3、C4、CO2和H2S）浓度（Vol、Inter）、C5相对分子质量分数和注入气体比重有关。ANN模型的一个主要优点是，对于复杂的油气组分，MMP值也可以达到较高精度。统计比较结果表明，所建立的人工神经网络模型比经验相关法取得了更好的预测结果。ANN模型的平均绝对百分比误差为13.46%，均方根误差为3.6，Pearson相关系数为0.95。敏感度分析表明，在建立合适的混相注入条件时，注入气体的比重和温度是最主要的考虑因素。在已有的关联式中，Yellig和Metcalfe关联式具有较好的预测性能，但其精度要低于ANN模型。

Estimation of minimum miscibility pressure of varied gas compositions and reservoir crude oil over a wide range of conditions using an artificial neural network model

Abouzar Choubineh, Abbas Helalizadeh, David A.Wood

(Published: 2018-10-03)

CorrespondingAuthor and Email:  David A.Wood, dw@dwasolutions.com; ORCID: https://orcid.org/0000-0003-3202-4069

---

Citation: Choubineh, A., Helalizadeh, A., Wood, D.A. Estimation of minimum miscibility pressure of varied gas compositions and reservoir crude oil over a wide range of conditions using an artificial neural network model. Advances in Geo-Energy Research, 2019, 3(1): 52-66, doi: 10.26804/ager.2019.01.04.

---

ArticleType: Original article

---

Abstract：

          Minimum miscibility pressure (MMP) is a key variable for monitoring miscibility between reservoir fluid and injection gas. Experimental and non-experimental methods are used to estimate MMP. Available miscibility correlations attempt to predict the minimum miscibility pressure for a specific type of gas. Here an artificial neural network (ANN) model is applied to a dataset involving 251 data records from around the world in a novel way to estimate the gas-crude oil MMP for a wide range of injected gases and crude oil compositions. This approach is relevant to sequestration projects in which injected gas compositions might vary significantly. The model is correlated with the reservoir temperature, concentrations of volatile (C1 and N2) and intermediate (C2, C3, C4, CO2 and H2S) fractions in the oil (Vol/Inter), C5+ molecular weight fractions in the oil and injected gas specific gravity. A key benefit of the ANN model is that MMP can be determined with reasonable accuracy for a wide range of oil and gas compositions. Statistical comparison of predictions shows that the developed ANN model yields better predictions than empirical-correlation methods. The ANN model predictions achieve a mean absolute percentage error of 13.46%, root mean square error of 3.6 and Pearson's correlation coefficient of 0.95. Sensitivity analysis reveals that injected gas specific gravity and temperature are the most important factors to consider when establishing appropriate miscible injection conditions. Among the available published correlations, the Yellig and Metcalfe correlation demonstrates good prediction performance, but it is not as accurate as the developed ANN model.

---

Keywords: Minimum miscibility pressure, miscibility correlations, artificial neural network, statistical accuracy, sensitivity analysis, enhanced oil recovery.