Deep-Learning Prediction of Flowing Bottomhole Pressure in Gas-Lifted Unconventional Wells

Abstract Accurate prediction of flowing bottomhole pressure (FBHP) is essential for effectively designing and managing artificial lift systems, particularly for wells utilizing gas lift. Conventional methods for FBHP estimation, including empirical correlations and mechanistic models, often prove inadequate for unconventional wells, being either overly simplified or computationally prohibitive. To address this challenge, we develop and evaluate several deep learning architectures for predicting FBHP in unconventional shale oil wells under gas-lift operations. A comprehensive dataset is compiled from 21 oil wells in the Texas Permian Basin Shale, incorporating readily available parameters such as well depth, operating valve depth, and production/injection data. The predictive capabilities of an artificial neural network (ANN), a long short-term memory (LSTM) network, a hybrid LSTM-ANN model, and a Transformer model are evaluated. Five of the 21 wells are used to evaluate the deep learning model performance. The analysis of the results revealed factors that influence the deep learning model's performance. The result showed that the Transformer model's prediction performance is most reliable under different scenarios and among the four deep learning models. Besides, the selection bias in the training set can significantly impact the model's prediction performance. Furthermore, our analysis reveals that hyperparameter tuning reduces residual errorsby refining model parameters, leading to architectures that better capture the training data patterns and deliver enhanced predictive accuracy. This work highlights the significant potential of advanced deep learning models as practical tools for optimizing gas-lift operations across a wide range of fluid and reservoir conditions.