Accurate prediction of steel transportation volume is crucial for logistics planning in steel enterprises. However, existing methods often overlook two key challenges: the uncertainty in processing orders due to dynamic prioritization and resource competition, and complex temporal dependencies across multi-stage production processes that drive output fluctuations. To address these issues, we propose a Priority-Aware Temporal Embedding and Dynamic Hypergraph Network for steel transportation volume prediction (PTE-DHNet), which introduces two key innovations. First, a Priority-aware Temporal Embedding Network jointly models historical processing order patterns across products and dynamic order adjustments caused by resource competition, learning the impact of production priorities on output variability through feature fusion. Second, a Transfer Perception Dynamic Hypergraph models the time-varying influence of upstream outputs on downstream processing steps, capturing both intra-step product category interactions and inter-step variations in production output dynamics. Extensive experiments on three real-world steel production datasets show that our method outperforms representative baseline models, achieving over 5.7% improvement in WMAPE.