Reservoir simulation

Transformer-Based Models Aid Prediction of Transient Production of Oil Wells

The authors of this paper propose a novel approach to data-driven modeling for transient production of oil wells.

sliding-window technique graph
Fig. 1—The sliding-window technique cuts the initial time series into fixed-length samples used afterward to train and test machine-learning algorithms. The upper graph shows an example of the downhole pressure behavior, while the lower shows one of the time-dependent governing parameters (e.g., oil flow rate).

The authors of the complete paper propose a novel approach to data-driven modeling of transient production of oil wells, applying transformer-based neural networks trained on multivariate time series composed of various parameters of oil wells measured during exploitation. By tuning machine-learning models for a single well (ignoring the effect of neighboring wells) on open-source field data sets, the authors of the paper demonstrate that the transformer-based method outperforms recurrent neural networks (RNNs) with long- and short-term memory (LSTM) and gated recurrent-unit cells in the forecasting of bottomhole pressure dynamics.


The authors apply a novel deep-learning algorithm called a transformer to build surrogate models for simulations of well performance. Transformer architecture initially was developed for natural-language processing problems. However, in recent years, researchers have adapted transformers for time-series forecasting.

Restricted Content
We're sorry, but this content is reserved for SPE Members. If you are a member, please sign in at the top of the page for access. If you are not a member and you find JPT content valuable, we encourage you to become a part of the SPE member community.