CNN Collaborates with Bi-LSTM for Sedimentary Microfacies Identification at the Front Edge of Continental Delta

Traditional manual identification of sedimentary microfacies relies heavily on experience, which is inefficient and highly subjective. We focus on the identification of sedimentary microfacies in continental deltaic front and proposes a deep learning model integrating a Convolutional Neural Network with a Bidirectional Long Short-Term Memory Network to enhance the efficiency and accuracy of microfacies identification. First, the logging data from continental deltaic front are preprocessed, followed by stratigraphic layering and sedimentary microfacies classification based on curve morphology, core data, and other geological indicators, thereby constructing a dataset for sedimentary microfacies identification in continental deltaic environments. Subsequently, a deep learning model combining CNN and Bi-LSTM is designed. Feature engineering is optimized by incorporating differential features, sliding window statistics, and frequency-domain features. During model training, geological constraint correction and dynamic class-weight adjustment strategies are introduced to improve the model’s robustness and generalization capability. The AUC of the proposed model on the test set reaches 0.92, which is significantly better than the traditional manual analysis methods. The results indicate the effectiveness of deep learning in identifying sedimentary microfacies in continental deltaic front. We provide an intelligent solution for sedimentary microfacies identification in continental deltas and offers a novel technical approach for reservoir prediction in oil and gas exploration, holding significant theoretical and practical implications.