Evolutionary Approach to Optimal Oil Skimmer Assignment for Oil Spill Response: A Case Study

Area of Science:

• Environmental Science
• Operations Research
• Computer Science

Background:

• Current oil skimmer assignments face inefficiencies in deployment and capacity utilization.
• South Korea's regulations necessitate careful planning for oil spill response operations.

Purpose of the Study:

• To develop and validate a genetic algorithm for optimizing oil skimmer assignments.
• To enhance the efficiency of oil spill response operations by minimizing work time and mobilized locations.
• To introduce a deep neural network-based surrogate model for faster optimization.

Main Methods:

• A genetic algorithm with a tailored repair operation for constrained assignments was developed.
• Simulation-based evaluation was used to ensure compliance with South Korean regulations.
• A deep neural network surrogate model was implemented to accelerate the optimization process.
• Scenario-based simulations mimicking real-world oil spills were conducted for validation.

Main Results:

• Optimized assignments reduced average work time and total skimmer capacity compared to current methods.
• The deep neural network surrogate model significantly improved computational efficiency over simulation-based optimization.
• Minimizing mobilized locations led to substantial reductions in required deployment sites.
• The strategy demonstrated significant reductions in work time and required locations for major oil spills in South Korea.

Conclusions:

• The proposed genetic algorithm and mobilized location minimization strategy effectively enhance oil spill response operations.
• The study highlights the potential for computational optimization in environmental emergency management.
• Integration of deep learning models offers a promising avenue for accelerating complex simulation-based optimization problems.