基于RBF神经网络模型的天然气储层测录井岩性识别方法  

Lithological identification method for natural gas reservoirs based on the RBF neural network model using logging data

作　　者：冯昊楠 杨森[3] 王涛[3] 王依言 曾星海 王方正 FENG Haonan;YANG Sen;WANG Tao;WANG Yiyan;ZENG Xinghai;WANG Fangzheng(School of Earth Science and Engineering of Xi′an Shiyou University,Xi′an,Shaanxi 710065,China;Shaanxi Provincial Key Laboratory of Petroleum Accumulation Geology,Xi′an,Shaanxi 710065,China;No.3 Gas Production Plant of PetroChina Changqing Oilfield Company,ordos,Inner Mongolia 017300,China;No.11 Oil Production Plant of PetroChina Changqing Oilfield Company,Qingyang,Gansu 745000,China)

机构地区：[1]西安石油大学地球科学与工程学院 [2]陕西省油气成藏地质学重点实验室 [3]中国石油长庆油田公司第三采气厂 [4]中国石油长庆油田公司第十一采油厂

出　　处：《录井工程》2025年第1期63-69,76,共8页Mud Logging Engineering

摘　　要：准确确定储层中岩石的类型和特征,可以为天然气勘探和开发提供重要依据,为此研究基于RBF神经网络模型的天然气储层测录井岩性识别方法。该方法针对设备性能差异和人为操作误差等因素导致测录井数据存在的偏差,分别从数据参数转换、压力平衡和钻井液流速影响3方面对测录井数据进行实时校正预处理,运用经验模态分解方法对校正后的测录井数据进行分解,得到低频、高频测录井数据分量,并计算测录井数据不同分量的状态特征谱熵,再将其输入RBF神经网络模型内,经过模型传输、映射和分类,输出天然气储层的岩性识别结果。实验结果表明,该方法可有效对测录井数据实施校正预处理,同时能从测录井数据内分解其高频分量,可准确识别天然气储层的岩性,应用效果较为显著。Accurately determine the types and characterestics of reservoir rocks and provide an important basis for natural gas exploration and development,this paper,this study investigate a lithological identification method for natural gas reservoirs based on the RBF neural network model using logging data.In order to deal with such factors as equipment performance differences and human operation errors,this paper studies the lithology identification method of natural gas reservoirs based on logging data.After realtime correction and preprocessing of logging data from the three aspects of data parameter conversion,pressure balance,and drilling fluid flow rate,empirical mode decomposition method is used to decompose the corrected logging data,and obtain lowfrequency and highfrequency logging data components,and calculate the state characteristic spectral entropy of different components of logging data.Then,the state characteristic spectral entropy is input into the RBF neural network model.After model transmission,mapping,and classification,the lithology identification results of natural gas reservoirs is output.The experimental results show that this method can effectively correct and preprocess logging data,and meanwhile decompose its highfrequency components from logging data,and accurately identify the lithology of natural gas reservoirs.The application effect is significant.

关 键 词：测录井数据 天然气储层 岩性识别 校正预处理 经验模态分解 RBF神经网络 

分 类 号：TE132.1[石油与天然气工程—油气勘探]