世界核地质科学2025，Vol.42Issue(4)：782-792,11.DOI:10.3969/j.issn.1672-0636.2025.04.008
神经网络多属性融合技术定量预测砂体厚度和分布
Quantitative prediction of sandbody thickness and distribution via multi-attribute fusion technology in neural network
摘要
Abstract
As oil and gas exploration advances into the deep,the exploration difficulty continues to increase.Particularly for oil fields that have entered the mid-to-late development stages,the identification of thin interbedded sandstone bodies has become a core challenge.While seismic attributes can effectively characterize the lateral distribution of sand bodies,single attributes are insufficient for quantitatively predicting sandstone thickness,and serious multi-solutions exist between multiple attributes.To address this technical challenge,this study proposed a multi-attribute fusion method based on the Back Propagation(BP)neural network.The method selects root mean square amplitude,arc length attribute,and average envelope attribute to construct a nonlinear mapping model,tested with calibration data from 11 wells to practice the quantitative prediction of sandstone thickness.The results show that the correlation coefficient between predicted and actual sandstone thickness reaches 0.93,with errors controlled within 3 meters.The method effectively captures the spatial distribution of braided river,meandering river,and deltaic sandstone bodies in the study area,which will provide an efficient solution for predicting thin interbedded reservoirs in complex sedimentary environments.关键词
地震多属性融合/薄互层砂体/储层预测/BP神经网络/埕岛-桩海地区Key words
seismic multi-attribute fusion/thin inter-bedded sand body/reservoir prediction/BP neural network/Chengdao-Zhuanghai area分类
天文与地球科学引用本文复制引用
于世娜,杨健洪,周红科,张珈毓,董志文,武群虎,詹桐,彭定亮..神经网络多属性融合技术定量预测砂体厚度和分布[J].世界核地质科学,2025,42(4):782-792,11.基金项目
国家自然科学基金重点项目(编号:42230816)资助 Supported by Key Project of the National Natural Science Foundation of China(Grant No.42230816) （编号:42230816）
