Addressing the current issues of lacking standards and diverse user demands in smart home service management, this paper proposes a spatio-temporal data-driven control method for smart home service. The method involves constructing a temporal knowledge graph for smart home and utilizing a federated learningbased approach for smart home service management. By capturing the state of concept instances in smart home scenarios, the temporal knowledge graph provides temporal data on environmental changes and service statuses. Leveraging federated learning algorithms that amalgamate model parameters from various households enables personalized model updates and predictions of smart home service statuses. Experimental results demonstrate the method’s effectiveness in controlling smart home devices, accurately meeting user demands with high precision and rapid convergence speed.