In recent years, knowledge graphs (KGs) have gained significant attention in the field of recommender systems by incorporating various side information to enhance recommendation performance. To be more specific, KGs capture relationships among entities (e.g., items) and have been successfully utilized in various recommendation scenarios. However, previous studies have mainly focused on leveraging node information for making recommendations, disregarding the potential utility of relation paths within KGs. To address these limitations, this paper introduces a novel framework, called the Knowledge Graph Relation Patterns Network (KGRPN) for recommender systems. The proposed framework aims to incorporate user-specific interest patterns represented by sequential relation paths, thus providing valuable signals for capturing personalized preferences. The KGRPN framework employs a recursive method to encode user patterns derived from relational paths, enhancing the understanding of user-specific interests. Additionally, a novel scoring function is introduced to combine both node and relational information for calculating the matching score, enabling more accurate recommendation predictions. Our comprehensive experiments on real-world datasets demonstrate the effectiveness of the proposed KGRPN framework.