基于布谷鸟—BP神经网络的页岩脆性指数预测研究  

Prediction of shale brittleness index based on cuckoo-BP neural network

作　　者：黄开兴 刘卫华[2] 吴朝容[1] 胡华锋 周枫[2] 李勇[1] 陈朝譞 汪子祺 孙正星 Huang Kaixing;Liu Weihua;Wu Chaorong;Hu Huafeng;Zhou Feng;Li Yong;Chen Chaoxuan;Wang Ziqi;Sun Zhengxing(College of Geophysics,Chengdu University of Technology;Sinopec Petroleum Geophysical Technology Research Institute Co.,Ltd.,Sinopec Key Laboratory of Geophysics)

机构地区：[1]成都理工大学地球物理学院 [2]中国石化石油物探技术研究院有限公司中国石化地球物理重点实验室

出　　处：《中国石油勘探》2024年第2期158-166,共9页China Petroleum Exploration

基　　金：中国石化地球物理重点实验室开放基金资助项目“基于AI的页岩储层参数定量预测方法研究”(36750000-23-FW0399-0005)。

摘　　要：页岩储层具有低孔隙度、低渗透率的物理性质,因此在页岩气开采中往往需要对其储层进行压裂处理,而页岩储层的可压裂性可用脆性指数来评价。目前应用最广泛的岩石脆性指数计算方法是基于矿物组分法。基于矿物组分法计算获得岩心页岩脆性指数(BI),利用BP神经网络的自我学习能力,探寻测井参数与页岩脆性指数(BI)之间的非线性关系,再结合布谷鸟(CS)算法的全局优化能力和稳定性来提升BP神经网络的预测精度和稳定性,从而建立基于CS—BP神经网络的页岩脆性指数预测模型。使用CS—BP预测模型对研究区Y1井和Y2井两口井进行了页岩BI值预测,其预测结果显示:CS—BP预测值与岩心BI值的变化趋势基本一致;CS—BP预测值总体预测效果较好。研究结果表明:基于布谷鸟(CS)—BP神经网络,利用测井资料快速计算页岩脆性指数的方法在研究区具有一定的实用价值。Shale reservoir is characterized by physical properties of low porosity and low permeability,and fracturing treatment is often required in shale gas production.The fracability of shale reservoir can be evaluated by brittleness index(BI),and the mineral composition method is the most popular method for calculating rock BI.Based on the mineral composition method calculated core BI value,the non-linear relationship between logging parameters and shale BI has been analyzed by applying the self-learning ability of BP neural network,and then the cuckoo(CS)algorithm with global optimization ability and stability has been used to improve the prediction accuracy and stability of BP neural network,so as to establish a shale BI prediction model based on CS-BP neural network.The CS-BP prediction model has been used to predict shale BI values in wells Y1 and Y2 in the study area,which indicates that CS-BP prediction values have a basically consistent trend with core BI values,showing a good prediction result as a whole.The study results indicate that the cuckoo(CS)-BP neural network based method for rapidly calculating shale BI by using logging data has certain practical value in the study area.

关 键 词：可压裂性 页岩储层 BP神经网络 CS—BP算法 页岩脆性指数 

分 类 号：TE19[石油与天然气工程—油气勘探]