%0 Journal Article
%T 基于机器学习实现海上气田陆地终端液态产品产量预测与挖潜<br>The Prediction and Improvement of Liquid Hydrocarbons Production Based on the Machine Learning Algorithm
%A 羊新州
%A 闫正和
%A 罗睿乔
%A 杨鹏
%A 唐圣来
%J Journal of Oil and Gas Technology
%P 13-21
%@ 2471-7207
%D 2020
%I Hans Publishing
%R 10.12677/JOGT.2020.424107
%X 海上开采出来的天然气通过海底管线输送到陆地天然气处理终端，经过一系列处理工艺后，生产量丙烷、丁烷、轻烃和凝析油等副产品。副产品的产出不光与各气田本身的气质组分相关，同样受到陆地终端设备工况的影响。笔者首先通过分析陆地终端的工艺流程，归纳影响终端副产品的关键流程。然后将各类副产品的析出情况通过聚类分析，找出对各类副产品回收效率有影响的关键设备，筛选出相应的异常值，进行异常标注。再结合设备工况的标注信息，通过机器学习方法实现对液态产品产量的精准计算。最后挖掘生产潜力，预测在各设备完好条件下各液态产品的产量，为工艺流程的优化方向提供基础。<br />
The natural gas extracted from the offshore gasfield is transported to the onshore gas treatment terminal through the submarine pipeline. After a series of treatment processes, propane, butane, light hydrocarbon and condensate are produced. All these liquid hydrocarbons production is not only related to the temperament component of each gas field, but also affected by the working con-dition of the land terminal equipment. Firstly, by analyzing the technological process of the land terminal, the precipitation of all kinds of hydrocarbons production was analyzed through clustering to find out the key equipment that had an impact on the recovery efficiency. Accurate calculation of liquid hydrocarbons production output is achieved by machine learning method. Finally, using the machine learning model, we can predict the production of each liquid hydrocarbon under the different working conditions, and provide the basis for the optimization direction of the process flow.
%K 天然气，终端液态产品，异常标记，机器学习，潜力挖掘<br>Gas
%K Terminal Gas Byproduct
%K Abnormal Detection
%K Machine Learning
%K Prediction and Improvement
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=40169