A Data-Driven Approach for Stylolite Detection

Abstract

Abstract Stylolite is a specific geopattern that can occur in both sedimentary rocks and deformed zones, which could change porosity of the reservoir, modify the permeability, and even result in horizontal permeability barriers. Though there are many related studies to characterize this issue, most of them focused on experimental methods. In this work, we proposed a new approach for recovering geometrical information of the stylolite zone (including its size and location) based on neural network architectures including convolutional neural network (CNN), recurrent neural network (RNN) and attention, which could serve as a preliminary data-driven solution to the problem. To simplify the problem, we first conducted simulation by building three-dimensional multilayer reservoir models with one stylolite zone. We considered both simplified cases with only a few homogeneous layers, and cases with heterogeneous layers to generalize our work. For the heterogeneous case, we extracted the permeability from SPE10 model 2, a commonly used public resource. Producing and observation wells in the model are at different locations and provide pressure and production rate data as inputs for the deep learning models, in the form of multivariant time series data. For homogeneous cases, after zero-padding and standardizing our inputs to tackle different-length data and features with different scales, we passed our dataset to a CNN-LSTM model. The two subnetworks are connected in parallel to combine the advantages of CNN in extracting local temporal features and the strengths of LSTM in capturing long-time dependency via self-loops. Models containing only a few CNN and LSTM models are also covered in our work as baseline models. For heterogeneous cases, a CNN-based model U-net and an attention-based model SeFT were introduced to enhance the performance. On the homogeneous dataset, our CNN-LSTM model achieved a satisfactory performance and could predict the locations and sizes of the stylolite zone and outperformed the two baseline models. On the more challenging heterogeneous dataset, our baseline and CNN- LSTM models failed to deliver meaningful results. In contrast, SeFT and U-net showed success in the sense that we could successfully predict the locations of the stylolite zones.