Physics-Guided Deep Learning for Prediction of Energy Production from Geothermal Reservoirs

This study proposes a novel physics-guided recurrent neural network (RNN) model for optimizing geothermal reservoir management. Unlike traditional data-driven approaches, which often struggle with limited data and produce physically inconsistent predictions, this model incorporates the structure of physics-based equations directly into the RNN architecture. This approach improves prediction accuracy and generalization in optimization tasks, where well control strategies may extend beyond training data. We validate the model's performance using both numerical examples and real-world datasets.

📄 Download paper here