The integration of recommendation systems with knowledge graphs to perform hybrid recommendations has significant research value and implications. Addressing the technological service needs of Qinghai, we have researched and established a method for technological service recommendations. By harnessing the distinct characteristics of Qinghai's technological data and the recommendation requirements for technological services, we introduced the Adjust-RippleNet algorithm, building upon the foundational RippleNet recommendation algorithm. This method replaces the user's historical behavior data in the original algorithm with the most recent background data of entity objects. The accumulated latent preferences of entity objects across different ripple sets are then represented as vectors of the entity objects, forming an entity object-background interaction matrix. Ultimately, based on the labels of entities in the knowledge graph, we executed categorized recommendations and confirmed the efficacy of this algorithm through comparative experiments.