Production Forecasting Based on Attribute-Augmented Spatiotemporal Graph Convolutional Network for a Typical Carbonate Reservoir in the Middle East

Production forecasting plays an important role in development plans during the entire period of petroleum exploration and development. Artificial intelligence has been extensively investigated recent years because its capacity to analyze interpret complex data. With emergence spatiotemporal models that can integrate graph convolutional networks (GCN) recurrent neural (RNN), it is now possible achieve multi-well production prediction by considering impact interactions between producers historical data simultaneously. Moreover, accurate not only depends on but also influence neighboring injectors’ gas injection rate (GIR). Therefore, based assumption introducing GIR enhance accuracy, this paper proposes a deep learning-based hybrid model aimed at both characteristics injectors. Specifically, we integrated into attribute-augmented network (AST-GCN) gated units (GRU) extract intricate temporal correlations from The method proposed successfully applied well pattern (including five seven injectors) low-permeability carbonate reservoir Middle East. In single forecasting, error AST-GCN 63.2%, 37.3%, 16.1% lower MedAE, MAE, RMSE compared with GRU 62.9%, 44.6%, 28.9% RNS. Similarly, accuracy 15.9% 35.8% higher than RNS prediction. well-pattern 41.2%, 64.2%, 75.2% RMSE, MedAE RNS, while 29.3% higher. After different degrees Gaussian noise are added actual data, average change 3.3%, 0.4%, 1.2% which indicates robustness model. results show consider spatial correlation same time, performs oil forecasts.