Effective storm surge risk assessment and deep reinforcement learning based evacuation planning: A case study of Daya Bay Petrochemical Industrial Zone

Abstract

Storm surge is one of the most destructive marine disasters, characterized by abnormal and temporary rises in water levels during intense storms, leading to extreme inland flooding in the coastal area. storm surge risk assessment and evacuation planning, play a crucial role in saving lives and mitigating disasters. Conventional risk assessment struggles to meet the demands of refined risk evaluation research for small-scale elements, such as roads, and current evacuation plans are generally based on broader regional scales, failing to provide effective road-level evacuation planning for evacuees. This study developed five typical typhoon scenarios for the coupled ADCIRC-SWAN model to simulate storm surge inundation. Combining these simulations with road network, storm surge risk assessment was conducted in the Daya Bay Petrochemical Industrial Zone, a vulnerable low-lying coastal region of Huizhou City, China. Based on the risk assessment, a combination of the Deep Q-Network (DQN) model and raster environment was employed to develop real-time evacuation plans during storm surge events. To address the DQN model's convergence challenges, compressed search space and navigational reward methods were proposed. 1000 starting points were randomly selected for route planning, and the results indicate that the proposed method is highly effective in devising optimal evacuation routes with minimal deviation, offering valuable guidance for evacuees during real-world storm surges.