Abstract
As a key clean energy impacting the global energy structure, the efficient exploration of shale gas holds significant importance. An efficient development approach can significantly reduce environmental pollution caused during shale gas extraction. In this research, we present an intelligent method for shale gas well trajectory identification based on a deep learning framework. Utilising element logging data, the approach employs BiLSTM, multiple attention mechanisms, and PINN constraint modules to achieve accurate well trajectory recognition despite highly imbalanced data. Final results demonstrate that this method elevates formation identification accuracy to 96.4%, with minority class prediction accuracy exceeding 90%. Guided by BiLSTM and attention mechanism, the model acquires bidirectional reception and processing capabilities for stratigraphic information, enabling well trajectory identification and prediction through globally global optimization. Concurrently, the model clearly shows the importance of various element features in classification decisions and the influence of feature attributes at different depths on final outcomes. Within the PINN architecture, 84.5% of erroneous results deviating from geological and geophysical principles were effectively corrected. In summary, this method not only substantially improves the accuracy of shale gas well trajectory identification to enhance the efficiency of shale gas resource exploration and reduce the pollution underground, but also provides a new paradigm for the application of deep learning technologies in the geo-energy field.
Biography
Dr Wang Qiaochu, Lecturer at the College of Geosciences, China University of Petroleum (Beijing). His primary research focuses on artificial intelligence and oil and gas exploration, concentrating on the deep integration of AI technologies with petroleum exploration and development, alongside environmental protection. He has undertaken 12 high-level research projects, including the National Natural Science Foundation of China and American Association of Petroleum Geologists, and has published over 40 SCI-indexed papers.
