Knowledge, as the core driving force of intelligent applications, directly impacts the reliability and effectiveness of downstream tasks such as recommendation system and wind energy forecast through its completeness. Real-world knowledge evolves over time, and Temporal Knowledge Graph Completion (TKGC) aims to complete missing entities or relations under temporal constraints to enhance the completeness of knowledge. However, Existing research mostly focuses on the local temporal dependencies of entities, neglecting the combined influence of different time domains during their long-term evolution processes. To address this challenge, We propose a dual-module-driven completion model. The model employs a dual-module joint mechanism to capture the time-dependent features of relations and entities during evolution. Specifically, the evolutionary feature decomposition module utilizes the temporal decomposition capability of time series decomposition methods to decompose the temporal evolution of relations into periodic and trend components, enabling explicit modeling for precise perception of relational evolutionary features. The global correlation optimization module compresses global temporal information using discrete cosine transform to extract cross-temporal association features. It further integrates entity information to dynamically adjust and optimize the current temporal context, thereby capturing cross-temporal entity correlations. Extensive experiments conducted on four public temporal knowledge graph datasets demonstrate the effectiveness of the proposed models in enhancing temporal knowledge completion performance.