摘要
针对高含水期水淹层变化的动态特性,提出一种基于主成分分析的最小二乘支持向量机水淹层动态预测方法.该方法应用数据挖掘方法与改进的支持向量机方法,研究高含水期水淹层的分类识别问题,找到测井参数曲线与水淹级别之间的非线性映射关系,建立适合高含水期水淹特征的动态识别模型.它不仅充分考虑各种影响因素,而且利用主成分分析法准确提取影响水淹级别划分的测井参数曲线,避免模型输入参数间存在相关性导致划分精度低以及模型求解复杂、训练速度慢的缺点.结果表明,该方法较其他方法具有更快的运算速度和更高的识别符合率,其运算速度为43s,识别符合率达到97.0%,能体现高含水油田水淹层的动态变化特征.
According to the characteristic of water-flooded zone during high water cut stage,dynamic prediction method of water-flooded layer with least squares support vector machine based on the principal component analysis is proposed in this paper.This method used data mining method and the improved support vector machine method to study the classification of water-flooded layer in high water cut oilfields;found the non-linear mapping between logging curve and water-flooding levels;and established the dynamic recognition model of water-flooded layer log interpretation.It not only fully considers the various influence factors,but also extracts logging parameter curve that can affect the division of water flooded levels by using the principal component analysis method,which can avoid the shortcomings of low accuracy of division for there are correlation between input parameters and solving complex and slow training when solving the model.The results show that the operation speeds and recognition precision of the method proposed in this paper are all better than other recognition methods;its average running time is only 43 seconds,average recognition precision is 92%.Moreover the new method could reflect the dynamic characteristics of water-flooded layer in high water cut oil fields.
出处
《大庆石油学院学报》
CAS
北大核心
2011年第2期51-55,118,共5页
Journal of Daqing Petroleum Institute
基金
四川省教育厅重点项目资助(07ZA143)
关键词
动态预测
水淹层识别
主成分分析
最小二乘支持向量机
dynamic prediction
water-flooded layer recognition
principal component analysis
least squares support vector machine