Conditioning of deep-learning surrogate models to image data with application to reservoir characterization
作者：
Highlights：
• We proposed a knowledge-based hybrid workflow for reservoir heterogeneity characterization.
• The state-of-ther-art residual-in-residual dense block is employed to address highly-complex non-Gaussian models.
• Results show that a significant reduction in computational cost was achieved while the accuracy remains on geological parameter estimations.
• The hybrid workflow is generic and should be applicable in many geophysical applications as well as other engineering domains.
摘要
•We proposed a knowledge-based hybrid workflow for reservoir heterogeneity characterization.•The state-of-ther-art residual-in-residual dense block is employed to address highly-complex non-Gaussian models.•Results show that a significant reduction in computational cost was achieved while the accuracy remains on geological parameter estimations.•The hybrid workflow is generic and should be applicable in many geophysical applications as well as other engineering domains.
论文关键词：Reservoir simulation,Data parameterization,Deep convolutional neural network,Image segmentation,Data assimilation
论文评审过程：Received 14 July 2020, Revised 3 February 2021, Accepted 13 March 2021, Available online 16 March 2021, Version of Record 22 March 2021.
论文官网地址：https://doi.org/10.1016/j.knosys.2021.106956
