面向复杂储层的非均衡测井数据的岩性识别  

Lithology identification for imbalanced logging data on complex reservoirs

作　　者：瞿晓婷 张蕾[1] 冯宏伟[1] 王惠亚[2] 张涛[3] 冯筠[1] 

机构地区：[1]西北大学信息科学与技术学院,西安710127 [2]西北大学数学学院,西安710127 [3]西北大学地质系,西安710127

出　　处：《地球物理学进展》2016年第5期2128-2132,共5页Progress in Geophysics

基　　金：国家自然科学基金项目;陕西省自然科学基金项目(2014JM8338);陕西省教育厅科学研究计划自然科学专项项目(15JK1738)联合资助

摘　　要：针对常规岩性识别方法在复杂储层的非均衡测井数据准确率不高的问题,本文提出了一种面向复杂储层非均衡测井数据的多层BP神经网络的岩性识别算法.根据测井数据的特点及测井数据间的非线性关系,对测井数据进行均衡化处理,利用多层BP神经网络进行逐层岩性识别.在13个含油气盆地的7793个地层元素测井数据中测试表明,本文提出的方法在数据非均衡比达到1∶6的情况下,岩性识别的符合率仍能达到92%以上,在复杂储层的岩性识别方面具有很好的应用前景.Traditional lithology identification method for imbalanced logging data of complex reservoirs often suffer from low accuracy. In this paper, we propose a multi-layer BP neural network specially for imbalanced logging data on complex reservoirs. Firstly, according to the characteristics and the non-linear mapping of the logging data, equalization processing is performed, and then multi-layer BP neural network is employed to do the lithology identification in the way of layer by layer. The experimental results in 7793 ECS data of 9 Oil and gas basins demonstrated that the average compliance rates of the proposed methods achieve 92% even in 1 ： 6 imbalanced ratio of the logging data. In the future, the proposed algorithm has a good prospect and give the valuable data for basis geology research work.

关 键 词：岩性识别 非均衡测井数据 多层BP神经网络 复杂储层 

分 类 号：P631[天文地球—地质矿产勘探]