Abstract: The current mainstream method of entity alignment in knowledge graph is to learn the embedding representation of knowledge graph through graph neural network and measure the similarity between entity embeddings to achieve entity alignment. Many entity alignment methods only consider the structure information and relation information of knowledge graphs, while attribute information is often ignored. Aiming at the above problems, this paper proposes an entity alignment method for fusing attribute embeddings: Relation-aware Dual-Graph Lite Convolutional Network fusing Attribute, RDGLite-A. This method firstly extracts the relation information of the knowledge graph based on the attention mechanism, and then uses the graph convolutional network with highway to obtain attribute information, lastly integrates the embeddings information of the two to achieve higher-accuracy entity alignment. The experiment results on three cross-lingual datasets show that this method enhances the entity representation ability by fusing attribute information in knowledge graph. On 3 datasets, compared with the original model, the Hits@1 values have increased by 6.42%, 4.59% and 1.98%, respectively, and the alignment performance is significantly better than the current mainstream entity alignment methods.