基于SVM模型的排水采气效果预测研究
Research on Prediction of Drainage Gas Production Effect Based on SVM Model
陆国琛 1葛俊瑞 2秦丙林 1郭志辉 1田天 1王颖1
作者信息
- 1. 中国石油化工股份有限公司上海海洋油气分公司石油工程技术研究院,上海 200120
- 2. 中海石油(中国)有限公司上海分公司,上海 200335
- 折叠
摘要
为解决目前排水采气工艺效果预测不佳的问题,以涡流排水采气为例,采用SVM(support vector machine)模型进行效果预测,充分考虑地层压力、油藏中深、孔隙度、渗透率、产水量、产气量、气水比等影响涡流排采效果的因素,收集 7个区块涡流排水采气影响因素参数作为输入向量,平均单井日增气量和平均单井日增水量作为输出向量,通过样本训练确定SVM模型最优惩罚参数C为 22.1453,最优的γ为 0.30885,控制误差为 0.01.利用建立的SVM模型对T气田C1、C8井涡流排采生产情况进行预测,并与实际生产情况进行对比,结果表明平均单井日增气量相对误差为 8.57%,平均单井日增水量相对误差为 7%.研究表明基于SVM涡流排水采气效果预测模型预测值与实际值接近,模型预测结果准确可靠.
Abstract
In order to solve the problem of poor prediction effect of the current drainage gas production process,taking the eddy current drainage gas production as an example,the SVM(support vector machine)model is adopted to predict the effect,fully considering the formation pressure,reservoir depth,porosity,permeability,water production,gas production,gas water ratio and other factors that affect the effect of eddy current drainage gas production,and collecting the parameters of the influencing factors of eddy current drainage gas production in seven blocks as the input vector.The average daily gas increase of single well and the daily water increase of single well are taken as output vectors.Through sample training,the optimal penalty parameter C of SVM model is determined to be 22.1453 and the optimal γ is 0.30885,of which the control error is 0.01.The established SVM model is used to predict the production of eddy current drainage in Well C1 and C8 of T gas field,and the results are compared with the actual production.The results show that the relative error of the average daily gas increase of a single well is 8.57%,and the relative error of the average daily water increase of a single well is 7%.The research shows that the prediction value of the eddy current drainage gas production effect based on SVM model is close to the actual value,and the prediction result of the model is accurate and reliable.
关键词
SVM模型/排水采气/效果预测/气田Key words
support vector machines model/drainage gas production/effect prediction/gas field引用本文复制引用
出版年
2024
NETL
NSTL
万方数据