基于近似支持向量机的流体识别方法  

A Fluid Identification Method Based on Proximal Support Vector Machine

作　　者：刘佳乐[1] 文晓涛[1] 张瑞[1] 李天[1] 李世凯[1] 

机构地区：[1]成都理工大学地球探测与信息技术教育部重点实验室,四川成都610059

出　　处：《辽宁化工》2016年第4期449-453,共5页Liaoning Chemical Industry

摘　　要：多属性分析是支持向量机的一个重要功能。它可以基于同一目标的多个属性进行综合分析并得出反映该事物属性特征的最佳判别规则,是解决识别问题的一种重要方法。相对于传统的支持向量机,近似支持向量机(PSVM)在大数据量训练集的处理效率上具有较明显的优势且准确度高,适合处理信息量丰富的测井属性组合。一般情况下,采用两种属性值做交会图并通过数据的分布差异来定性区分流体或岩性的做法较为常见,也可利用拟合等方式来实现简单的定量划分。但这些方式均在精度上有所欠缺,且不能正确划分相互重合区域的两类数据。本文讨论了基于近似支持向量机的流体识别方法。通过利用流体识别因子或各测井曲线值作为训练集,并求解基于等式约束的最优化问题得到最优决策函数,可在短时间内获取属性值分布上的学习经验。该方法在数据点相互重合区域的识别效果上有一定的改善。将该方法用于某地龙王庙组地层的气水识别,能够实现较高的准确率。Multi-attribute analysis is an important function of support vector machine. It can analyze several different feature attributes of a target together, and obtain an optimal discrimination rule which is related to the target＇s freature attributes. Relative to the conventional support vector machines, the proximal support vector machine has advantage in efficiency. PSVM can process large dataset very quickly, and it is suitable for processing log character attributes combination which contains abundant data. Typically, we usually use two different attributes to form a cross plot. By the distribution difference of the data, the fluids or lithology can be distinguished qualitatively. The fitting method can also distinguish fluids or lithology approximately in a quantitative way. But these methods both have less accuracy, and they can＇t distribute the two types of data points which located in the corss-section. In this article, a logging identification and analysis method based on support vector machine was discussed. Using fluid identification factor or other kinds of logging curve values as the data set, an optimal decision function can be obtained by solving equality constrained optimization problem. Then the PSVM can get the learning experience of attributes＇ location. This method can do better in distributing some data points which located in the corss-section. By this method, the distributing of gas or water of Longwangmiao formation reservoirs can show a higher accuracy.

关 键 词：测井分析 近似支持向量机 特征属性 流体 识别 

分 类 号：P631.4[天文地球—地质矿产勘探]