基于粒子群优化支持向量机的注水井吸水剖面预测  

Prediction of water injection profile based on particle swarm optimization-support vector machine

作　　者：李俊键[1] 周代余[2] 赵冀[2] 丁帅伟[1] 姜汉桥[1] 张曙振[2] 梁彬 旷曦域[2] 王倩[2] 

机构地区：[1]中国石油大学(北京)石油工程教育部重点实验室,北京102249 [2]中国石油塔里木油田分公司勘探开发研究院,新疆库尔勒841000

出　　处：《中国海上油气》2016年第5期66-70,共5页China Offshore Oil and Gas

基　　金：国家青年自然科学基金"复杂裂缝潜山油藏水平井堵水封堵机理研究(编号:51404280)"部分研究成果

摘　　要：注水井吸水剖面预测对于分层注水调整方案的编制和注采井组水淹状况的确定具有重要的指导意义。历史吸水剖面是目前最准确的计算分层吸水量的资料,但是目前对于该资料的使用并不是很全面,多因吸水剖面资料过少,或插值方法选取过于简单。本文在历史吸水剖面资料的基础上,利用粒子群优化支持向量机方法,通过回归拟合历史吸水剖面资料,建立了吸水剖面预测模型,实现了无吸水剖面时间点注水井吸水剖面的预测,从而达到准确劈分注水量的目的。实例应用结果表明本文方法考虑了各种影响小层吸水的动静态因素,预测结果准确可靠,可用于注水井分层注水量的劈分研究。The prediction of water injection profile is important for guiding the adjustment scheme of layered water injection and determination of water flooding status. Historical injection profile is the most accurate material to calculate the layered water absorbing capacity, but it is not used comprehensively because of the rare material or too simple interpolation method. In this paper, based on historical injection profile data, the injection profile prediction model is established through the regression fitting history injection profile data based on particle swarm optimization-support vector machine method（PSO-SVM）, thus the prediction of injection profile without data is realized. The application results show the proposed method is accurate and reliable, which can provide the foundation for the division of injected-water volume for water injected well.

关 键 词：吸水剖面预测 支持向量机 粒子群算法 回归预测模型 

分 类 号：TE319[石油与天然气工程—油气田开发工程]