当前位置: X-MOL 学术J. Pet. Sci. Eng. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Accurate Determination of Natural Gas Compressibility Factor by Measuring Temperature, Pressure and Joule-Thomson Coefficient: Artificial Neural Network Approach
Journal of Petroleum Science and Engineering Pub Date : 2021-01-21 , DOI: 10.1016/j.petrol.2021.108427
Mahmood Farzaneh-Gord , Hamid Reza Rahbari , Behnam Mohseni-Gharesafa , Alexander Toikka , Irina Zvereva

Natural Gas (NG) compressibility factor as important property at any NG industrial applications determined by utilizing an intelligent approach precisely. Three thermodynamic properties include pressure, temperature and Joule-Thomson (JT) coefficient are selected as input parameters. These properties are chosen due to the measurement capabilities of available sensors. Unlike the traditional approaches, the current approach does not require NG compositions as input. The current intelligent approach is developed based on an Artificial Neural Network (ANN) method. Real-time measurement capability and very low cost are two main advantages of the developed approach. Big data sets of NG thermodynamic properties are created considering 30,000 random compositions for training, testing and validating the ANN. The GERG-2008 is utilized (as the most recent equation of state) to calculate thermodynamic properties to train the ANN. Validation of the developed ANN method compared to experimental data shows the Average Absolute Percent Deviation (AAPD) is about 0.33%. To show the accuracy of the developed approach, four different NG compositions are selected as case studies. The compressibility factor and JT coefficient are computed for various pressure and temperature range using the traditional approach. Then, the compressibility factor is determined using the intelligent approach when only pressure, temperature and JT coefficient are known. The AAPD of NG compressibility factor calculations for various natural gases show 0.385% for pure methane, 0.45% for the Khangiran gas, 0.58 % for the Kangan gas, 0.78% for the Pars gas and is 1.12% for the Bidboland gas. The comparing results show that overall AAPD is less than 0.7% that shows the high accuracy of the intelligent approach.



中文翻译:

通过测量温度,压力和焦耳-汤姆森系数精确确定天然气可压缩系数:人工神经网络方法

天然气(NG)的可压缩系数是在任何NG工业应用中通过精确利用智能方法确定的重要属性。选择三个压力,温度和焦耳-汤姆森(JT)系数作为输入参数。选择这些属性的原因是可用传感器的测量能力。与传统方法不同,当前方法不需要NG成分作为输入。当前的智能方法是基于人工神经网络(ANN)方法开发的。实时测量能力和极低的成本是该开发方法的两个主要优点。考虑到30,000种随机成分用于训练,测试和验证ANN,创建了NG热力学性质的大数据集。使用GERG-2008(作为最新的状态方程)来计算热力学特性以训练ANN。与实验数据相比,已开发的ANN方法的验证显示平均绝对偏差(AAPD)约为0.33%。为了显示所开发方法的准确性,选择了四种不同的天然气成分作为案例研究。使用传统方法可针对各种压力和温度范围计算可压缩系数和JT系数。然后,当仅知道压力,温度和JT系数时,使用智能方法确定可压缩系数。各种天然气的NG压缩系数的AAPD计算值显示,纯甲烷为0.385%,康加兰气为0.45%,康安气为0.58%,帕尔斯气为0.78%,为1。Bidboland气体的12%。比较结果表明,总体AAPD小于0.7%,表明该智能方法具有很高的准确性。

更新日期:2021-01-22
down
wechat
bug