In the field of representation learning on knowledge graphs (KGs), a hyper-relational fact consists of a main triple and several auxiliary attribute value descriptions, which is considered to be more comprehensive and specific than a triple-based fact. However, the existing hyper-relational KG embedding methods in a single view are limited in application due to weakening the hierarchical structure representing the affiliation between entities. To break this limitation, we propose a dual-view hyper-relational KG (DH-KG) structure which contains a hyper-relational instance view for entities and a hyper-relational ontology view for concepts abstracted hierarchically from entities to jointly model hyper-relational and hierarchical information. In this paper, we first define link prediction and entity typing tasks on DH-KG and construct two DH-KG datasets, JW44K-6K extracted from Wikidata and HTDM based on medical data. Furthermore, We propose a DH-KG embedding model DHGE, based on GRAN encoder, HGNN, and joint learning. Experimental results show that DHGE outperforms baseline models on DH-KG. We also provide an example of the application of this technology in the field of hypertension medication. Our model and datasets are publicly available.