基于隐马尔科夫模型的复杂碳酸盐岩岩性识别——以苏里格气田苏东41-33区块下古气藏为例  

Complex Carbonate Lithology Recognition Using the Hidden Markov Model with Well Logging Data: Case study in the Sutong 41-33 block of Sulige gas field

作　　者：赵默雷 段正军[2] 高世臣 田苗 

机构地区：[1]中国地质大学(北京)数理学院,北京100083 [2]中国地质大学(北京)能源学院,北京100083 [3]中国人民大学统计学院,北京100083

出　　处：《数学的实践与认识》2017年第24期118-126,共9页Mathematics in Practice and Theory

摘　　要：测井数据是进行岩性分类的最主要资料,但因为碳酸盐岩低孔,各向异性,非均质性的特点,造成石灰岩、白云岩和其过渡岩性的测井资料内在属性复杂,单从测井资料的数据分布结构较难解决岩性分类问题.为了降低岩性反演的多解性,需要对岩性资料进行空间相关性的研究.研究表明,采用隐马尔科夫模型,开展以转移概率为基础的垂向空间岩性序列展布规律探索,利用垂向展布规律作为约束,结合动态求解策略,能够提高岩性分类的预测精度.使用该模型对苏里格气田苏东41-33区块下古气藏复杂碳酸盐岩进行岩性识别,和传统统计学方法高斯混合模型相比,该方法能提高预测准确度5.89%以上.同时,该模型能够揭示某种沉积环境,避免最终的预测结果产生实际工区未曾出现的岩性序列.Logging parameters are the most important data for lithologic classification, but because of the low porosity, anisotropy and heterogeneity of carbonate rocks, the multiple solutions of logging parameters of limestone, dolomite and its transition lithology, Making it more difficult to solve the problem of lithology classification from the attribute structure of logging parameters alone. As a kind of spatial data, the study of the spatial correlation of lithologic data can reduce the multiplicity of lithological prediction and improve the prediction accuracy. The study shows that the study of distribution law of lithologic sequence in vertical space based on transfer probability is carried out by using hidden Markov model. By using the vertical distribution law as constraint and the dynamic and efficient solution of the model,Classification prediction accuracy. This model can be used to identify the lithology of complex carbonate rocks in Su-41-33 block of Sulige gas field. Compared with traditional statistical methods, this method can improve the prediction accuracy by more than 5.89%. The use of this model can reveal a depositional environment and avoid the resulting lithologic sequence that is not present in the actual work area.

关 键 词：复杂碳酸盐岩岩性识别 隐马尔科夫模型 岩性序列垂向展布 转移概率矩阵 

分 类 号：P618.13[天文地球—矿床学]