4 Presentations

• 
  03:30 PM - 03:55 PM

  Optimizing Pallet Transloading Operations in a port terminal facility

  • Noam Mssellati, presenter, Polytechnique Montréal
  • Daniel Aloise, Polytechnique Montréal
  • Simon Boivin, Vooban

  Transloading is a supply chain operation that involves transferring goods from one mode of transport to another with minimum storage and delay. It helps optimize shipment costs and is often necessary when cargo is exported. The challenges in transloading are similar to cross-docking or transshipment. While previous research has focused mainly on scheduling and strategic decision-making, few studies address facility operational planning. This work optimizes the operational costs of a pallet transloading facility with a fixed shipment arrival order, taking into consideration special mixing constraints and storage capacity limitations. A two-step resolution approach is adopted. First, we enumerate decision variables, from a customized resource graph, representing each a consolidation plan for a sequence of batch arrivals. Then, we solve a Mixed-Integer Linear Program (MILP) that integrates these variables into a single optimization model that searches for the best subset of consolidation plans. Our approach is compared against the industry-standard First In, First Out (FIFO) strategy. Computational experiments demonstrate that it achieves cost-efficient solutions within practical resolution times.

• 
  03:55 PM - 04:20 PM

  A Joint Pricing and Investment Optimization Problem for Port Network

  • Xiaoyue Zhu, presenter, HEC Montreal
  • Yossiri Adulyasak, HEC Montréal
  • Wenyi Xia, HEC Montréal

  Pricing and capacity investment are critical strategic decisions for port authorities operating in competitive markets. This paper presents a bilevel optimization model to jointly optimize pricing and capacity investment in a port network, where the port authority (leader) determines the pricing and capacity investment levels for the ports under its control, while shippers (followers) adjust freight flows in response to the leader’s decisions. To address this problem, we first implement a KKT-based single-level reformulation strategy and deal with multiple nonlinear terms. Subsequently, we develop an exact decomposition method to solve the bilevel problem. Numerical results on the performance of the models will be presented. Finally, we will reveal some managerial insights about port pricing and investment decisions.

• 
  04:20 PM - 04:45 PM

  Realtime convex optimization for hybrid ship optimal operation

  • Louis-Philippe Baillargeon, presenter, Polytechnique Montréal
  • Berger Maxime, Université du Québec à Rimouski
  • Antoine Lesage-Landry, Polytechnique Montréal

  In this work, we present a unified optimization framework for shipboard power systems that simultaneously determines the vessel’s trajectory, speed profile, and power management strategy. Our approach integrates modern power sources, e.g., solar photovoltaics, onshore power connections, batteries , in addition to the ship’s diesel generators while accounting for the physical constraints of both the ship’s electrical network and its itinerary. The problem is formulated as a mixed-integer convex program and can be solved to global optimality with off-the-shelf solvers while considering all operational decisions. To optimize both path and speed while avoiding obstacles, e.g., land, shallow waters or protected areas, the non-convex navigable space is decomposed into a set of convex, obstacle-free regions, which can then be embedded in the optimization problem using disjunctive constraints. The convex regions also serve to evaluate the impact of spatially varying meteorological conditions like wind and wave on energy efficiency. The proposed method is validated on a test case involving a voyage along the Saint-Laurent River, to be completed in a predetermined amount of time. The system comprises six electrical buses and traverses five convex regions.

• 
  04:45 PM - 05:10 PM

  Machine Learning Algorithms for Joint Prediction of Ship Berthing Times and Tugboat Allocation: A Case Study at the Port of Quebec

  • Marie-Sabine Saget, presenter, FSA-Laval University/CIRRELT
  • Maryam Darvish, Université Laval
  • Jacques Renaud, Université Laval, CIRRELT

  This study presents a machine learning (ML) framework for predicting ship docking times and optimizing tugboat allocation using 2021 data from the Port of Quebec. Docking times are forecasted using Elastic-NET, Decision Tree, and Random Forest regression models, while tugboat allocations are determined through a heuristic approach, Naïve Bayes, Decision Tree, and Random Forest classifiers. For the ML methods, Bayesian optimization handles hyperparameter tuning, while SHapley Additive exPlanations or SHAP tests assess feature impacts.
  
  Results indicate that Random Forest models outperform other approaches in both tasks. The model's regressor achieves an R2 of 88% for docking time predictions, and the classifier attains accuracy, precision, recall, and F1-scores around 93% for tugboat allocation. Overall, the models' performance and interpretability highlight the framework's potential to enhance decision-making at the Port of Quebec, particularly in long-term forecasting and resource management.