基于集成机器学习模型的混合原油凝点预测方法  

Gel point estimation method of mixed crude oil based on ensemble machine learning model

作　　者：何宇轩 苏怀[1,2] 张成 苏杨 李鸿英 黄骞[1,3] 张劲军 HE Yuxuan;SU Huai;ZHANG Cheng;SU Yang;LI Hongying;HUANG Qian;ZHANG Jinjun(National Engineering Laboratory for Pipeline Safety in China University of Petroleum(Beijing),Beijing 102249,China;Key Laboratory of Beijing City for Urban Oil and Gas Transmission and Distribution Technology in China University of Petroleum(Beijing),Beijing 102249,China;PetroChina Planning and Engineering Institute,Beijing 100083,China)

机构地区：[1]中国石油大学(北京)油气管道输送安全国家工程实验室,北京102249 [2]中国石油大学(北京)城市油气输配技术北京市重点实验室,北京102249 [3]中国石油规划总院,北京100083

出　　处：《中国石油大学学报(自然科学版)》2025年第2期214-222,共9页Journal of China University of Petroleum(Edition of Natural Science)

基　　金：国家自然科学基金青年科学基金项目(51904316);中国石油大学(北京)科研基金项目(2462021YJRC013)。

摘　　要：混合输送是不同原油在同一管道中输送最常用的一种方式,快速、精准地掌握混合原油的流动参数,是制定混合原油配输方案、保证管道安全高效灵活运行的基础,通过人工取样测试确定混合原油凝点,难以及时有效对进管原油进行管控;运用基于组分原油配比和凝点的经验模型计算混合原油凝点,虽简便易行,但在方法上存在预测精度提升的瓶颈;建立一种基于XGBoost集成机器学习模型的混合原油凝点预测方法。结果表明:当模型以组分原油凝点、密度、黏度以及配比为输入参数时,经8912组数据训练后的混合原油凝点预测平均绝对偏差为1.12℃;当输入参数中组分原油凝点缺失时,预测平均绝对偏差为1.93℃,其中绝对偏差小于2℃的占88.0%。Mixed transport is the most common way to transport multiple crude oil in the same pipeline.Grasping the flow properties of the mixed oil quickly and accurately is the basis of making the mixed crude oil distribution scheme and ensuring the safe,efficient and flexible operation of the pipeline.The gel point of mixed crude oil is often determined by the manual sampling test,so it is difficult to effectively control the crude oil into the pipeline in time.It is simple and easy to calculate the gel point of mixed crude oil by using the empirical model based on the ratio and gel point of component crude oil,but there is a bottleneck in the method to improve the prediction accuracy.An integrated machine learning model based on XGBoost was proposed to predict the gel point of mixed crude oil.The results show that,with the inputs of gel point,density,viscosity and ratio in component oils,the mean absolute error of the model prediction estimations after training with 8912 data is 1.12℃.When the gel point of the component crude oil is missing,the mean absolute error is 1.93℃and the percentage of the predicted absolute error within 2℃is 88.0%.

关 键 词：混合原油 凝点 机器学习 预测 

分 类 号：TE832[石油与天然气工程—油气储运工程]