Research on Optimization of Rail Transport Delay Cost Based on SA-LNS

To address the key engineering problem of cascading delay of physical bottleneck sections during peak operation, this paper proposes a hybrid SA-LNS algorithm to optimize the delay cost of railway transportation. By integrating Large Neighborhood Search’s destruction-repair mechanism with Simulated Annealing’s probabilistic acceptance criterion, the algorithm establishes a closed-loop perturbation-screening-convergence framework. It dynamically optimizes train arrival/departure timings and dwell strategies while satisfying safety headways, minimum station dwell times, and inter-station energy constraints. Validated on Lanzhou Metro Line 1’s Yellow River Bridge bottleneck—representative of capacity-constrained railway sections—the method achieves a 24.1% reduction in bottleneck delay costs and converges to a minimum total cost of 219,000 within 128 iterations. Computational efficiency exceeds standalone SA by 97.39% and LNS by 32.63%, enabling real-time applications. This approach effectively resolves network-wide delay propagation and provides rapid timetable recovery under disruptions, demonstrating significant value for high-density rail systems where scheduling inefficiencies in constrained infrastructure elevate operational costs. The algorithm’s balanced optimization of delay penalties, rescheduling expenses, and operational expenditures offers a systematic solution for metropolitan rail networks with complex spatial-temporal constraints.