Transfer learning–based seismic identification of deep small-scale strike-slip faults in the central Sichuan Basin (SW China)

Résumé fourni par la source

Deep (>4500 m) fault-related reservoir has become an important target for hydrocarbon exploration and development. However, accurate characterization of small-scale faults in deep subsurface remains highly challenging. For this contribution, the transfer learning-based seismic data processing was carried out in the central Sichuan Basin, northwest China. It has shown that many more small-scale strike-slip faults are identified and automatic mapped. This method has the advantage than conventional seismic methods with weak seismic responses and low signal-to-noise ratios. Furthermore, transfer learning can address the limitations of conventional deep learning method that needs many fault labels and synthetic data for pre-training. The results are confirmed by horizontal wells across the strike-slip fault zones that achieved much higher gas production from the deep tight reservoirs. This case study highlights that transfer learning-based seismic processing can significantly improves seismic prediction of deep strike-slip faults.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.