Energy-Aware Operations for Decarbonizing Ports and Shipping under Uncertainty

Abstract

Maritime decarbonization is transforming ports and shipping networks into increasingly interconnected operational and energy systems. However, operational planning, energy management, demand prediction, and uncertainty handling are often studied separately, limiting their applicability to electrified ports and emerging alternative-fuel networks. This thesis develops a layered framework for energy-aware operations planning across port and maritime systems under uncertainty, combining stochastic optimization, machine learning, and distributionally robust optimization.

At the container-terminal level, a two-stage stochastic mixed-integer programming model coordinates vessel scheduling, cargo-handling equipment, and energy management under uncertain vessel arrivals and electricity prices. A progressive hedging algorithm is developed to solve the resulting problem. Experiments based on Container Terminal Altenwerder in Hamburg show that demand-responsive planning can shift flexible electricity loads and reduce operational costs by up to 13.2% under real-time electricity pricing.

At the berth level, a data-driven predict-then-optimize framework constructs time-granular shore power demand profiles from vessel port-call records and static vessel characteristics. Machine-learning models are used to predict these profiles under different levels of information availability, after which the predictions are integrated into a shore-power-aware berth allocation model. Experiments using data from Hamburg and Hong Kong demonstrate that prediction-informed planning substantially reduces peak-capacity violations and produces solutions close to the perfect-information benchmark. The results also show that the model with the highest predictive accuracy does not necessarily yield the best downstream operational decisions, underscoring the importance of evaluating prediction and optimization jointly.

At the shipping-network level, deterministic, chance-constrained, and Wasserstein-based distributionally robust models jointly optimize sailing speed, bunkering locations and quantities, and fuel choice for dual-fuel liner services under uncertain port fuel supply. The distributionally robust approach substantially improves out-of-sample feasibility, although at moderately higher planned costs. Methanol-uptake experiments further reveal that bunkering-station coverage determines the operational potential for alternative-fuel use, while price competitiveness determines how closely that potential is realized.

Overall, this thesis demonstrates that embedding energy considerations and uncertainty management directly into operational planning can improve economic efficiency, energy utilization, and operational reliability. It contributes an integrated operations-research and data-driven perspective for supporting the transition toward lower-emission port and maritime transport systems.