The shipping industry faces a pressing challenge with carbon emissions, prompting a focus on speed optimization for energy conservation and emission reduction. While much research has centered on optimizing speeds in oceans and rivers, canals have received less attention, despite their unique challenges of narrow waterways and busy locks. This study fills this gap by establishing a fuel consumption prediction model integrating key environmental factors such as water depth, width, and flow velocity. Drawing upon established methodologies in speed optimization, this study augments these models with waiting time limits for each canal segment. To validate the efficacy of the model, three representative ships are selected as case studies. The findings reveal a high predictive capability of the fuel consumption model, as evidenced by R² values exceeding 0.97 across all cases. Notably, the optimization approach yields a fuel consumption reduction ranging from 4% to 5% for short waiting times. Furthermore, compared to conventional methods, the proposed optimization strategy achieves an 8.19% enhancement in fuel consumption and carbon emission reduction for long waiting times, culminating in an overall optimization rate of 11.54%. These results underscore the significance of employing the proposed speed optimization methodology, particularly during peak periods of canal congestion.

This paper addresses the bidirectional multi-chamber lock scheduling problem by formulating a multi-objective mixed-integer linear programming (MILP) model that simultaneously minimizes average ship waiting time and maximizes chamber utilization. A tailored adaptive large neighborhood search (ALNS) algorithm is developed specifically based on the principle of the destruction and reconstruction of solutions. The algorithm efficacy is validated using the real-word data from Huai’an Lock of the Subei canal. The scheduling rules and parameters are defined from practical operation records. Simulation results demonstrate that the ALNS-based optimization significantly improves lock performance with average chamber utilization increasing by 12.98% and waiting time decreasing by 44.40%. Sensitivity analyses on objective weights further confirm the robustness of the proposed method. Benchmark comparisons with a greedy heuristic, genetic algorithm (GA), and particle swarm optimization (PSO) highlight the effectiveness and computational efficiency of ALNS. This study further explores a threshold-based directional control strategy, showing that relaxing strict alternating-direction rules under asymmetric traffic demand can improve efficiency. The findings provide practical insights for lock scheduling, offering decision support for lock authorities in designing adaptive scheduling and directional control policies.