Knowledge graph completion (KGC) aims to enhance the completeness and utility of knowledge graphs (KGs) by predicting and filling missing information. Existing methods primarily focus on structured representation learning, extracting low-dimensional embeddings of entities and relations to uncover and predict missing information in knowledge graphs. However, these methods often overlook entity type information and lack deep feature extraction capabilities. Inability to recognize the type information of entities may lead to poor embedding expression effects of entities, while insufficient deep feature extraction limits the model’s ability to understand complex relationships. To address these issues, this paper proposes a Knowledge Graph Completion with Entity Type-Aware and Deep Feature Extraction (TAFE). The model employs a Type-Aware Graph Attention Encoder (TA-GAT) to identify equivalence relations and model entity type information during graph context entity aggregation. Additionally, it incorporates a Deep Feature Extraction 3D Convolution Decoder (FE-Conv3D), using Gaussian function mapping techniques to capture deep feature information of entities and relations. The 3D convolutional kernels extract interaction and local features among embeddings, enhancing the model’s ability to capture details and understand complex relationships. Extensive experimental analysis demonstrates the effectiveness of TAFE in knowledge graph completion tasks.