Abstract: Demand uncertainty and inland port congestion pose significant challenges to the efficient operation of port collection and distribution systems. This paper proposes a bi-level programming model that simultaneously account for demand uncertainty and inland port congestion. The upper-level model optimizes inland port location and capacity configuration by minimizing construction costs and congestion costs,while the lower-level model optimizes transport paths by minimizing transportation costs,transfer costs,carbon emission costs,and time costs. A Genetic Algorithm is employed to solve the model,with empirical analysis based on data from the Liaoning port cluster and its northeastern hinterland. The results show that higher-grade inland ports effectively absorb demand fluctuations and reduce congestion costs,while reducing carbon emissions by increasing the share of railway transport. A maximum regret value of θ = 0.25 achieves the optimal balance between total cost and stability of the Liaoning network. Additionally,a carbon tax of ￥0.5/kg achieves the best emission reduction and congestion control outcomes.