<p>Accurate and rapid traffic flow prediction plays a crucial role in the application and development of Intelligent Transportation System(ITS). Existing methods typically rely on static graph structures, placing greater emphasis on employing complex models to capture temporal correlations, spatial dependencies, and spatiotemporal data heterogeneity, while often neglecting the potential impact of different features on traffic flow prediction. Consequently, these approaches struggle to capture real-time traffic pattern changes and overlook the integration of multi-period features (e.g., daily/weekly periodicity) and external factors such as weather conditions. To address these limitations, this study proposes a novel dynamic fusion graph convolutional traffic flow prediction framework with external factors and multi-period features DGCP-net. First, during the stage of data preprocessing, Adaptive Dynamic Fusion Graph Generation Module (ADFG) is introduced to adaptively construct time-varying adjacency matrices and dynamic correlation matrices through data-driven learning, overcoming the limitations of predefined static graphs. Second, Gated Fusion Graph Convolutional Unit(GFGCU) employs dual-gated recurrent mechanism combined with dynamic fusion graph convolution to simultaneously extract spatiotemporal features. Third, daily and weekly periodic feature matrices are embedded into the ADFG module to model the periodic patterns of traffic flow. Finally, feature fusion layer is incorporated to integrate external factors such as weather conditions, thereby enhancing prediction robustness. Experiments based on real-world traffic datasets PeMS and weather records demonstrate that the proposed model, compared to existing baseline methods, achieves improvements in three evaluation metrics(RMSE, MAE, MAPE) across four datasets.</p>