Abstract: Cardinality estimation in dynamic knowledge graphs is crucial for optimizing queries in graph databases. The task goal is to obtain accurate query cardinality in dynamic graphs for input query statements and then select efficient execution plans to optimize the query process through cost analysis. Existing solutions mainly focus on static graphs and cannot solve the problem of estimation accuracy on large knowledge graphs that are constantly growing. The proposed cardinality estimation model based on lifelong learning of knowledge graph embedding incrementally acquires the changing query graph, uses lifelong embedding learning to create meaningful embeddings for dynamic entity relationships in knowledge graphs, captures the correlation between connections between different types of query graphs and data graphs, and then uses graph neural networks to learn query representations and estimate the cardinality of connected queries. The model can be applied to growing knowledge graphs and more accurately predict query cardinality in dynamically changing knowledge graph scenarios.