Seismic lithofacies prediction via deep learning with prior geological constraints

Reliable lithofacies prediction from seismic data is essential for advancing exploration and characterization in hydrocarbon reservoirs, CO2 storage, geothermal energy recovery, and groundwater management. Lithofacies prediction from seismic data under a data-driven supervised deep learning framework is commonly subject to issues of geological prior inconsistency, poor generalizability, and weak interpretability. To address these challenges, we propose the multi-seismic with multi-geological constraint Net (MSMGNet), a deep learning-based lithofacies prediction approach that not only incorporates multiseismic information but also takes into account the prior geological knowledge including stratigraphic differences, prior lithologic probabilities, lithofacies transition probabilities, and thickness distributions of the lithofacies. The input-based strategy and loss-based strategy are proposed to incorporate statistical geological features for constraining lithofacies prediction. In the input-based approach, geological features are embedded as structured input vectors, where stratigraphic units are encoded using one-hot encoding, and lithofacies proportions, transition probabilities, and thickness distributions are vectorized into numerical feature representations. In the loss-based approach, these geological features are formulated as regularization terms within the loss function, where Kullback–Leibler (KL) divergence is used to penalize deviations between predicted and prior distributions. Cross-well blind tests in a complex coal-bearing clastic reservoir demonstrate that incorporating these statistical geological features significantly improves the lithofacies prediction performance. More importantly, geology-constrained deep learning approaches enhance the ability to capture lithological variations between stratigraphic units, characterize thickness distributions, and identifying thin sandstone and coal layers. The input-based MSMGNet, which integrates all four geological constraints, achieves the highest prediction accuracy and geological consistency in comparison with the baseline model without geological constraints and the loss-based MSMGNet.