Rethinking intermodal hinterland freight transport: the hidden impact of demurrage, detention, and storage costs

This study investigates the critical role of demurrage, detention, and storage (DDS) costs in shaping transport decisions within hinterland intermodal networks, where such indirect charges are often overlooked in traditional logistics models. To address this limitation, we develop a DDS-integrated mixed-integer optimization model that jointly minimizes transportation and indirect costs while enforcing strict delivery deadlines. A preprocessing phase eliminates infeasible or dominated routes. The model is reformulated as a multi-commodity minimum-cost flow (MCMCF) framework where containers with different deadlines are treated as separate commodities. To overcome the computational complexity of the reformulated model, we propose the Marginal Cost Hybrid with Machine Learning (MCHML) algorithm. This approach integrates a four-layer architecture consisting of an initial cost-driven allocation stage, an XGBoost marginal cost estimator for guiding reassignment, a knapsack-based mechanism for resolving capacity overloads, and a simulated annealing (SA) refinement to escape local optima. These machine-learning modules provide predictive guidance that enhances the effectiveness and scalability of heuristic search. Benchmarking further indicates that MCHML achieves faster runtimes than column-and-row (C&R) generation and consistently outperforms the Adaptive Large Neighborhood Search (ALNS) heuristic in both solution quality and runtime. Using operational data from the Port of Le Havre, France, the results show that incorporating DDS costs reduces total logistics costs by 34%–45.3%, lowers CO2 emissions by 53.3%–64.7%, and increases intermodal share by 58.2%–81.8%. These findings challenge the perception that DDS costs inherently favor direct trucking and demonstrate the value of DDS-aware, learning-enhanced optimization for improving the economic and environmental performance of intermodal freight systems.