特高含水期油田产量预测新方法  

A new method to predict production of oilfields in ultrahigh water-cut stage

作　　者：钟仪华[1] 张志银[1,2] 朱海双[1,3] 

机构地区：[1]西南石油大学理学院,四川成都610500 [2]中联重科土方机械分公司,陕西渭南714000 [3]中联煤层气有限责任公司晋城分公司,山西晋城048200

出　　处：《断块油气田》2011年第5期641-644,共4页Fault-Block Oil & Gas Field

基　　金：四川省高等学校科技创新重大培育项目"油气田开发系统预测与优化控制的理论及应用"(09ZZ035);西南石油大学自然科学基金项目"智能预测方法及其在油田开发中的应用研究"(2010XJZ195)

摘　　要：特高含水期是油田重要的开采阶段,我国东部大部分油田已经进入特高含水期,研究处于该阶段的油田产量预测问题具有重要意义。由于处于特高含水期阶段的油田一般采集到的开发指标和影响因素的样本数较小,所以基于统计学理论的常规预测方法都不太适合该阶段的产量预测。文中从数据挖掘和机器学习角度提出了油田产量预测的新方法——基于主成分分析和支持向量机(SVM)方法的油田产量预测模型。先把由影响产量的众多因素构成的高维向量通过主成分分析技术进行数据降维后作为SVM的输入,通过SVM训练得到模型,并利用遗传算法(GA)优选模型的参数建立特高含水期油田产量预测模型。与其他方法相比,该方法减少了模型输入变量的维数,提高了模型收敛的速度和预测精度,能较好地反映特高含水期油田产量的动态变化规律。Because ultrahigh water-cut stage is an important period in oilfield development and most oilfields in eastern China have entered this development stage, it is significant to research the production prediction of oilfields in this stage. When an oilfield has been in this stage, the numbers of samples collected for oilfield development indices and their influence factors are small, many traditional predicting methods based on statistics theory are not really suitable for production prediction of oilfields in this stage. This paper proposes a new method to predict the production of oilfield from the point of data mining and machine learning, that is the prediction model of an oilfield production based on principal component analysis（PCA）and support vector （SVM） theory. At first, it reduces the dimension of the high dimension variable which consists of many influence factors to a lower dimension variable through PCA technology, then inputs the new lower dimension variable to SVM and establishes an oilfield production prediction model whose parameters are selected by genetic algorithm（GA）. Compared with other methods, this method can reduce the dimension of model input variables, improve the speed of model convergence and the precision of prediction. It also can reflect the dynamic laws of oilfield production in ultrahigh water-cut stage.

关 键 词：产量预测 支持向量机 主成分分析 遗传算法 特高含水期 

分 类 号：TE323[石油与天然气工程—油气田开发工程]