摘要
横波速度是岩性预测与流体识别的重要参数,对弹性参数反演和储层精细刻画具有非常重要的价值.但是,由于勘探成本控制,地球物理测井通常较少开展横波直接测量,而一般是通过经验公式或岩石物理模型实现横波速度预测.东海西湖凹陷油气储层为富含薄煤层的砂泥薄互层,储层岩石矿物含量与孔隙结构复杂,纵向上岩性和物性变化快,用经验公式或岩石物理模型预测横波速度精度不高.由于深度学习具有强大的复杂非线性表达和特征提取能力,为此利用深度学习方法开展横波速度预测研究.选取储层深度、纵波速度、密度、自然伽马、泥质含量、孔隙度等6种测井特征参数作为标签数据,构建了特征参数与横波速度之间复杂非线性映射的深度神经网络模型,对东海西湖凹陷地区富煤砂泥薄互层复杂储层进行了横波速度预测,实际测井数据的训练及测试结果表明,深度学习网络能够对东海西湖凹陷富煤砂泥薄互层进行高精度横波预测.
Shearing wave velocity is a crucial parameter for lithology prediction and fluid identification,playing a significant role in elastic parameter inversion and detailed reservoir characterization.However,due to constraints in exploration costs,direct shearing wave velocity measurements through well-logging are relatively rare and are typically estimated using empirical formulas or rock physics models.The hydrocarbon reservoirs in the Xihu Sag of the East China Sea are characterized by thinly interbedded sand-mud layers with abundant coal seams.The complex mineral composition and pore structure,along with rapid vertical variations in lithology and petrophysical properties,make it challenging to achieve high-accuracy shearing wave velocity predictions using empirical formulas or rock physics models,which often rely on simplified assumptions that fail to capture the complexity of the subsurface.Deep learning,with its powerful nonlinear representation and feature extraction capabilities,can effectively learn the intricate relationships between logging parameters and shearing wave velocity.In this study,a feedforward deep learning artificial neural network is employed to predict shearing wave velocity.Six logging parameters including reservoir depth,P-wave velocity,density,natural gamma,shale content,and porosity are used as input features to construct the deep learning neural model.A multi-layer network is designed to establish a complex nonlinear mapping between these parameters and shearing wave velocity,and an appropriate optimization strategy is implemented for artificial neural model training.This study focuses on shearing velocity prediction in the Xihu Sag of the East China Sea,where the reservoir is characterized by interbedded sandstone and mudstone with thin coal seams.The complex structural features and rapid vertical lithological variations present significant challenges for accurate shearing wave velocity estimation.Training and testing with actual well log data demonstrate that the feedforward deep learning neural network enables high-precision shearing wave velocity prediction for the complex thinly interbedded sand-mud reservoirs with coal enrichment in this region.
作者
郭咏欣
李键
秦德文
麻纪强
吴树梁
耿建华
GUO YongXin;LI Jian;QIN DeWen;MA JiQiang;WU ShuLiang;GENG JianHua(School of Ocean and Earth Sciences,Tongji University,Shanghai 200092,China;Zhonghai Petroleum(China)Co.,Ltd.Shanghai Branch,Shanghai 200335,China)
出处
《地球物理学进展》
北大核心
2025年第5期2148-2159,共12页
Progress in Geophysics
关键词
西湖凹陷
富煤地层
砂泥薄互层
深度学习
横波预测
Xihu Sag
Coal-rich formation
Sand-mud thin interbed
Deep learning
Shearing wave velocity prediction
