The Entity alignment aims to align entities across different Knowledge Graphs and plays an important role in knowledge fusion tasks. In recent years, existing methods have typically used multiple aspects of information to enhance the accuracy of matching, which makes the matching process more comprehensive. However, existing methods that integrate multiple features have relied on fixed weighting strategies, making it challenging to adjust the relative importance of each feature. Therefore, this paper proposes an entity alignment framework(HGEA) based on highway gates dynamic adaptive network. Firstly, HGEA aggregates the structure, name, and node importance of entities. Among them, to fully utilize of the structural information of the Knowledge Graphs, we propose an entity and relation representation strategy based on a graph attention network. Then, we have a highway gates dynamic adaptation network, which combines the gating mechanism and multi-level self-attention to dynamically adjust the weight of each feature to adapt to the information flow of different entity features and optimize the entity alignment. Experiments on three real-world cross-lingual datasets demonstrate that our framework surpasses the current state-of-the-art methods.