摘要
气井产能预测是气藏开发过程中的重要工作之一,在气田的整体评价和高效开发进程中具有很强的预见性。而对于"三超"气井来说,进行产能测试面临着较大困难,因此寻找一种较为方便的产能预测方法尤为重要。基于此,研究以气井静态资料、探井资料为依据,建立了多元线性回归、BP神经网络、支持向量机3种预测模型,通过对上述3种产能模型预测结果及3种预测方法的优缺点综合对比分析可知,基于支持向量机的气井产能预测模型预测精度较高、预测结果稳定、可操作性强,是一种适合库车前陆盆地"三超"气井产能预测的数据建模方法。
For most gas wells, to predict their productivity is very critical in reservoir engineering. Moreover, it plays a strong predictable role in whole evaluation and highly efficient development.However, it's very hard to predict the productivity of 3-ultra gas wells. So, it's critical to find out some convenient and suitable prediction methods. In this study, based on some static data and other data on exploration wells, we developed three predictable models including multivariable linear regression, BP neural network, and support vector machine(SVM), and comprehensively contrasted their advantages and disadvantages, and predicted results. It is shown that, characterized by higher accuracy and easily operability,the SVM-based model with more stable results deserves a suitable method for 3-ultra gas wells in Kuche Foreland Basin.
出处
《天然气技术与经济》
2018年第2期31-34,共4页
Natural Gas Technology and Economy
基金
中国石油科技创新基金项目(2015D-5006-0207)
重庆科技学院研究生科技创新计划项目(YKJCX2014020)
关键词
“三超”气井
产能预测
多元线性回归
BP神经网络
支持向量机
3-ultra gas wells
productivity prediction
multivariable linear regression
BP neural network
support vector machine