Driven by the carbon-neutrality target, port transportation requires routing methods that jointly optimize economic efficiency, carbon emissions, and network robustness. Existing studies mainly rely on static carbon constraints and conventional evolutionary strategies, which often suffer from poor adaptability and limited Pareto diversity in large-scale multimodal transport networks. To address these issues, this paper proposes an enhanced low-carbon port path optimization framework (LCPPO-E) based on an improved NSGA-III. The main algorithmic innovations include a congestion-aware dual-layer chromosome encoding mechanism, an adaptive reference-point update strategy guided by population distribution, and an external elite archive with self-tuning crossover-mutation coordination. A tri-objective optimization model is constructed to minimize total transport cost, carbon emission intensity, and node congestion risk simultaneously. Shanghai Port container transport data are used for validation. Experimental results show that, compared with the traditional cost-minimization model, the proposed method reduces carbon emission intensity by 21.3% and node congestion rate by 28.6%, while increasing total cost by only 5.1%. In comparison with GA, PSO, NSGA-II, and standard NSGA-III, the proposed algorithm achieves the best comprehensive performance, with a hypervolume improvement of 6.1% over standard NSGA-III and an increase in multimodal transport share to 41.8%. The results verify the effectiveness of the proposed method in solving low-carbon and resilient port routing problems.