2017年6月第36卷第3期
大庆石油地质与开发
Petroleum Geology and Oilfield Development in Daqing
D0I;10.3969/J.ISSN.1000-3754.2017.03.023
基于 GA-SVR的 CO，
驱原油
最小混相压力预测模型孙雷罗强潘毅冯洋
（西南石浦大学，四川咸都610500）
June,2017 Vol. 36 No. 3
摘要：为了得到更精确的CO，驱原油最小混相压力，考虑挥发组分（N,+CO,+CH，+H,S）含量、中间烃组分（C26）含量、重质组分（C）含量、重质组分的相对分子质量、重质组分密度以及温度的影响，建立了基于遗传算法参数寻优的支持向量回归机模型。模型优点在于使数据结构风险最小化，是基于数据精度高和回归函数复杂性适宜的条件下进行全局参数导优得到最优模型，根据测试样本数据可以给出预测结果，得到更为准确的最小混相压力数值。该模型计算结果平均相对误差为3.44%，与文献中的实验结果、细管实验结果对比，具有较好的准确性。
关键词：CO，题；最小混相压力；遗传算法；模型；支持向量回归机
文章编号：1000-3754（2017）03-0123-07
中图分类号：TE357
文献标识码：A
PREDICTINGMODELOFTHEOILMINIMALMISCIBLEPRESSURE
FORTHECO,FLOODINGBASEDONGA-SVR
SUN Lei,LUO Qiang，PANYi,FENGYang
(Southacest Petroleum University,Chengdu 610500,China)
Abstract: In order to more precisely obtain the oil minimal miscible pressure of CO, flooding, considering the in-fluences of the Use other people's result, study the various factors between contents of volatile ( N,+CO,+CH,+ H,S), intemediate hydrocarbon (C2-6), heavy hydrocarbon (C,), molecular weight (Mc,） and density (pc） of the heavy components and temperature, the optimized support vector regression (SVR) machine model was es-tablished on the basis of the genetic algorithm ( GA) parameters, The advantages of this model is to make the data structure risk minimal, which is the obtained optimized model from the overall parameters under the conditions of the high-precision data and suitable complexity of the regression function, and moreover with the help of the testing sample data, the predicted results were presented, thus the more accurate minimal miscible pressure value was ob tained. Comparing the model calculation results with the experimental ones in the references and slim-tube test, the average relative error is 3. 44% i. e. much better accuracy is presented.
Key words: carbon dioxide (CO,) flooding; minimum miscible pressure; genetic algorithm (GA); model; sup 收稿日期：2016-07-28
改回日期：2017-03-29
（01110）
作者简介：孙雷，男，1954年生，教授，从事油气相态理论与测试、气田与凝析气田开发设计及注气提高采收率方面
研究工作。
E-mail ; sunleiswpi @163. com
万方数据