Temporal knowledge graph reasoning aims to inferring missing facts within a temporal knowledge graph, which playing a crucial role in predicting time-evolving information in real-world context. Existing research predominantly emphasize the utilization of temporal dependencies of quadruples in graphs, often overlooking the evolution, transformation, or interaction of events over time, as well as the semantic information related to entities. Therefore, we propose a novel framework named Temp-EASE to incorporate both the evolution pattern of events and the semantic meaning of entities for temporal knowledge graph reasoning. Specifically, Temp-EASE treats each relation in the TKG as an event and constructs an event evolution graph to capture the evolution patterns of events over time. Furthermore, recognizing that the semantic content of entities in graphs may change over time, we design a template to prompt a large language model (LLM) for generating temporal background knowledge. Experiments on three benchmark datasets showcase that Temp-EASE outperforms state-of-the-art models, demonstrating its effectiveness in temporal knowledge graph reasoning.