A multi-objective optimization framework for urban flood mitigation using machine learning and optimization...

Area of Science:

• Environmental Engineering
• Computational Fluid Dynamics
• Machine Learning

Background:

• Urban flood mitigation design is critical due to climate change and urbanization.
• Traditional 1D hydrodynamic models lack accuracy for overland flow.
• 1D-2D models are accurate but computationally expensive for optimization.

Purpose of the Study:

• To develop a computationally efficient framework for optimal urban flood mitigation design.
• To integrate a machine learning surrogate with a multi-objective optimization algorithm.
• To overcome the computational limitations of high-fidelity hydrodynamic models in optimization.

Main Methods:

• Developed a multi-objective optimization framework.
• Employed a machine learning model as a surrogate for a 1D-2D hydrodynamic model.
• Validated the framework using a case study for urban flood mitigation.

Main Results:

• The machine learning surrogate accurately predicted inundation maps, matching high-fidelity models.
• Achieved a three-order-of-magnitude improvement in computational efficiency.
• The optimized scheme resulted in ¥113.5 million in life-cycle cost savings and enhanced flood protection.

Conclusions:

• The proposed framework provides a computationally tractable solution for optimal urban flood mitigation.
• Machine learning surrogates can bridge the gap between model fidelity and computational cost in optimization.
• This approach enables more effective and cost-efficient urban flood risk management.