Temporal Knowledge Graph Reasoning is an essential research area in the field of Temporal Knowledge Graphs, which make predictions about the future based on historical events. However, existing models have some drawbacks. (1) Current methods ignore influence of time information in historical snapshots. (2) Only the final prediction utilizes feature mining of temporal attributes. (3) The global information captured along the timeline is not complete enough. To address the aforementioned problems, we propose a temporal knowledge graph reasoning model named TIA-Net in this paper. TIA-Net is a convolution graph network model that integrates time-aware information using a multi-head attention mechanism. Firstly, the model incorporates time information encoding at each historical timestamp, temporal dependencies are captured from multiple perspectives using a multi-head attention mechanism. Secondly, the model utilizes a relational convolution graph network with evolutionary representation to capture structural dependencies among events in each snapshot. Lastly, the frequency characteristics of events, obtained by a global frequency adjustment mechanism, are used to improve prediction accuracy. Extensive experiments on five benchmark datasets show that TIA-Net outperforms the state-of-the-art temporal knowledge graph reasoning methods in most cases. Our source code is available at https://github.com/Cur-pro/TIA-Net .