Decision optimization for improved flood management of urban metro system

Abstract

In the context of increasing climate change and rapid urbanization, storm floods have become a significant threat to the operation of urban infrastructure. Among various infrastructure systems, metro systems are particularly at higher flood risk due to their underground location, confined spaces, and high population density. Recent metro flood incidents in cities like New York and Zhengzhou have caused substantial economic losses and even fatalities, highlighting the critical importance of effective flood management. However, the current metro flood management is mainly based on human experience, which may pose challenges to metro safety and compromise the performance of metro networks. Therefore, the purpose of this thesis is to develop novel and reliable frameworks that use optimization methods for decision-making in metro flood management, thereby enhancing the flood resilience of metro systems. To achieve the overall research objective, four major steps are proposed: first, assess and mitigate the flood risk using an evidential-reasoning-based approach; then, for station closure-protection issues, conduct pre-flood decision optimization and two-stage emergency optimization based on network topology; finally, optimize the location and inventory planning of flood control resource warehouses. The applicability and effectiveness of the proposed methodologies are verified on the Shanghai metro system in China. The main findings are summarized as follows: (1) The approach for flood risk assessment can provide reasonable, conservative, and discriminative assessment results, on which basis the sensitivity analysis can suggest effective risk mitigation measures. (2) For the pre-flood station closure-protection problem, the proposed optimization framework, which incorporates network topology, outperforms the experience-based baseline strategies in improving the overall performance of the metro network. (3) The two-stage stochastic optimization model offers refined closure schemes and dynamic, adaptive protection schemes for risky metro stations. Analysis using explainable artificial intelligence reveals that passenger flow and rainfall conditions in the sub-catchment area where the station is located are key factors influencing station closure. (4) The proposed multi-condition bilevel optimization method effectively determines warehouse locations, inventory, and resource transportation schemes during floods, thereby optimizing the performance of the metro network under various flood conditions.