In time-sensitive scenarios such as the integration of distributed renewable energy, strict requirements are imposed on the determinism of power communication networks. Although TSN (Time-Sensitive Network) technology provides deterministic guarantees within local area networks, existing architectures cannot ensure controllable end-to-end latency across wide areas. To address this, this paper designs a semi-distributed cross-domain TSN traffic scheduling architecture, in which a cross-domain coordinator selects paths for inter-domain traffic and distributes policies to intra-domain schedulers. Intra-domain schedulers then manage domain-specific traffic based on these policies. A traffic scheduling and routing model is developed, and an improved dual-color cooperative bat algorithm is designed, which incorporates a linkage update phase to control the global TSN traffic optimization direction, making the outcome as close as possible to that of centralized architectures. Experimental results show that the proposed method converges to an approximate optimal solution 26.5% faster than traditional algorithms and improves the scheduling success rate by 38.0% compared to traditional algorithms.