摘要
含气量是评价页岩气能否富集高产的主要参数之一,含气量越高越有利于页岩气井获得高产。传统含气量地震预测方法基于单属性、多属性的线性拟合或简单的神经网络,精度较低。基于深度神经网络的含气量预测方法,通过优选地震属性及最优化求解方法,选择合适的隐藏层个数、神经元个数、迭代次数来建立预测模型,从而预测页岩含气量,该方法能有效提高页岩含气量预测精度,为页岩气研究区地质评价、页岩气水平井井位布署提供支撑。
Gas content is one of the main parameters to evaluate whether shale gas can be enriched to obtain high-yield.The higher the gas content, the more favorable for shale gas wells to obtain high-yield.Traditional gas content seismic prediction methods are based on single-attribute, multi-attribute linear fitting or simple neural networks, and have low accuracy.The gas content prediction method is based on deep neural network.Through optimizing seismic attributes, optimizing the solution method and choosing the appropriate number of hidden layers, the number of neurons, and the number of iterations, a prediction model can be established to predict the gas content of shale, thus effectively improving the prediction accuracy of shale gas content and providing support for the geological evaluation of shale gas research areas and the deployment of shale gas horizontal wells.
作者
张勇
马晓东
李彦婧
蔡景顺
ZHANG Yong;MA Xiao-Dong;LI Yan-Jing;CAI Jing-Shun(Research Institute of Exploration and Development,East China Branch of SINOPEC,Nanjing 210005,China;Sichuan Baohua Xinsheng Oil&Gas Operation Service Co.Ltd.,Chengdu 610000,China)
出处
《物探与化探》
CAS
北大核心
2021年第3期569-575,共7页
Geophysical and Geochemical Exploration
基金
国家科技重大专项(2016ZX05061)
中国石油化工股份有限公司科技部项目(P19017⁃3)。
关键词
深度学习
页岩气
含气量
地震预测
非线性
deep learning
shale gas
shale gas content
seismic prediction
non-linear