Automatic identification system for shale faults and micro-fractures based on deep learning

Abstract

The present invention proposes an automatic identification system for shale faults and micro- fractures based on deep learning, belonging to the technical field of geological exploration. The system comprises a seismic data input unit, a preprocessing unit, a multi-scale feature extraction network, a fault and micro-fracture fusion identification unit, and a post-processing optimization unit. The multi-scale feature extraction network extracts large-scale fault features and small-scale micro-fracture features from pre-processed seismic data through parallel convolution branches; the fusion identification unit utilizes an attention mechanism to perform adaptive fusion on the multi-scale features, generating a joint identification map; the post-processing optimization unit performs topological continuity correction and false positive elimination on the map. The present invention realizes a full-process automated identification from seismic data to geological structure features, providing scientific and intelligent technical support for risk identification, well location optimization, and fracturing evaluation in shale gas exploration and development.