A Hybrid Tabular-Spatial-Temporal Model with 3D Geomodel for Production Prediction in Shale Gas Formations

Summary The evolution of shale gas production has reshaped North America’s energy profile. Using the vast amounts of data generated from production and operations, machine learning offers significant advantages in production forecasting and performance optimization. In this study, we propose a pioneering hybrid model that integrates tabular, spatial, and temporal modalities to enhance production forecasting in unconventional shale gas reservoirs. Despite traditional methods, such as artificial neural networks (ANN) and extreme gradient boosting (XGBoost), which rely solely on tabular data for training and prediction, we propose a novel 3D parameterization method. This approach tokenizes the formation property distribution into three-axis tensors, enabling a more comprehensive representation of spatial data. For this study, we established a 3D-convolutional neural network (3D-CNN) with an attention mechanism module to process the created spatial data. For temporal modality, we used the long short-term memory (LSTM) module to accept the dynamic input and predict the monthly production simultaneously. Data from a total of 677 wells in the Duvernay Formation were collected, preprocessed, and fed into the according module based on their modality. The results show that the model combining three modalities achieved an impressive level of accuracy, with a coefficient of determination (R2) of 0.8771, surpassing the tabular (0.7841) and tabular-spatial (0.8230) modality models. In addition, we applied global optimization to further enhance the model performance by optimizing the architecture of each module and model hyperparameters, and achieved a 1.88% improvement from the empirical design. These advancements set a new benchmark for predictive modeling in unconventional shale gas reservoirs, highlighting the importance of using data from different modalities in improving production forecast prediction.