A knowledge-based heterogeneity characterization framework for 3D steam-assisted gravity drainage reservoirs
作者：
Highlights：
• A hybrid deep learning-based workflow is developed to infer 3D shale heterogeneities.
• Novel inputs and outputs parameterization schemes are explored via wavelet transform.
• Deep learning is applied as a proxy of flow simulation to evaluate objective functions.
• Fast and satisfactory heterogeneities characterization in 3D reservoirs are obtained.
• Multiple characterized models exhibit similar shale distribution as the true model.
摘要
•A hybrid deep learning-based workflow is developed to infer 3D shale heterogeneities.•Novel inputs and outputs parameterization schemes are explored via wavelet transform.•Deep learning is applied as a proxy of flow simulation to evaluate objective functions.•Fast and satisfactory heterogeneities characterization in 3D reservoirs are obtained.•Multiple characterized models exhibit similar shale distribution as the true model.
论文关键词：SAGD,Deep learning,Convolutional neural network,Proxy model,Shale barriers,Heterogeneity modeling,Optimization
论文评审过程：Received 15 July 2019, Revised 29 November 2019, Accepted 30 November 2019, Available online 9 December 2019, Version of Record 24 February 2020.
论文官网地址：https://doi.org/10.1016/j.knosys.2019.105327
