A Machine Learning-based Hybrid Model for Fracture Parameterization and Distribution Prediction in Unconventional Reservoirs

Abstract

The understanding of fracture distributions plays a critical role in managing fractured reservoirs as they govern early water/CO2 breakthrough, impact sweep efficiency, and determine production behaviors. However, traditional simulation-based approaches, such as history matching, encounter significant difficulties in accurately predicting fracture distributions. Additionally, high-fidelity simulations can be computationally prohibitive. This paper proposes a comprehensive machine learning-based workflow to effectively characterize and describe the fracture distributions for unconventional reservoir models. The proposed workflow has four components. Firstly, a single fracture parameterization method is implemented, utilizing four fracture parameters: fracture initiation point, length, angle, and azimuth. Secondly, a Variational Autoencoder (VAE) model is employed for fracture map parameterization. The encoder maps a high-dimensional fracture distribution map to a low-dimensional latent space, and the decoder reconstructs the fracture distribution map from the reduced latent dimension to the full reservoir dimension. Thirdly, to predict the fracture distribution given only production data, a neural network is employed to establish a regression relationship between the latent variables and production data. Finally, to quantify reservoir model uncertainty, a nearest-neighbors selection is adopted by applying principal component analysis (PCA) in 2D principal. The efficacy of the proposed workflow is validated using a 2D synthetic case and a 3D benchmark case.