轨道交通网络韧性评估与最优恢复策略

王松，刘杰^*，黄健畅

山东交通学院交通与物流工程学院，山东 济南  250357

摘要：为提高轨道交通网络应对突发扰动事件的恢复能力，构建融合流量服务效率和网络运行效率的轨道交通网络韧性评估模型，以网络韧性最大化为目标建立轨道交通网络恢复策略模型，将遗传算法与自适应大邻域搜索算法相结合，提出一种混合自适应大邻域搜索遗传算法求解网络韧性最大化恢复策略模型；以杭州市轨道交通为实例，通过随机攻击和针对性攻击模拟自然灾害、人为破坏及运营管控等扰动情景，对比分析3种扰动情景下随机恢复策略、节点度优先恢复策略、重要度优先恢复策略和网络韧性最大化恢复策略的站点恢复顺序及网络韧性表现。结果表明：在3种扰动情景中，网络韧性最大化恢复策略对轨道交通网络的修复效果均最优，其次为节点度优先、重要度优先恢复策略，随机恢复策略的修复效果最差；人为破坏情景中，轨道交通网络所受影响最严重，此时网络韧性最大化恢复策略的网络韧性分别比其他3种策略分别增大6.7%、7.6%、25.1%；同时修复1个站点时，轨道交通网络约需8 h恢复至最优，同时修复4个站点时，约需3 h恢复至最优，可同时修复的站点越多，网络韧性恢复越快。在扰动发生初期尽快采用最优恢复策略并增大修复资源投入，可显著提高轨道交通网络的恢复效率与网络韧性。

关键词：轨道交通；网络韧性；恢复策略；遗传算法；自适应大邻域搜索算法

Resilience assessment and optimal recovery strategy of rail transit network

WANG Song, LIU Jie^*, HUANG Jianchang

School of Transportation and Logistics Engineering, Shandong Jiaotong University, Jinan 250357, China

Abstract: To improve the recovery capability of rail transit networks in response to sudden disturbance events, a resilience assessment model for rail transit networks is proposed, which integrates traffic service efficiency and network operational efficiency. A recovery strategy model for rail transit  is established with the objective of maximizing network resilience. By combining genetic algorithms with adaptive large neighborhood search algorithms, a hybrid adaptive large neighborhood search genetic algorithm is proposed to solve the resilience maximization recovery strategy model. Taking the rail transit system of Hangzhou as an example, disturbance scenarios such as natural disasters, human-made damages, and special control measures are simulated using random and deliberate attacks. The station recovery sequence and network resilience performance of random recovery strategies, node-degree-first recovery strategies, importance-first recovery strategies, and resilience-maximization recovery strategies are compared and analyzed under three disturbance scenarios. The results indicate that under all three disturbance scenarios, the resilience-maximization recovery strategy achieves the best repair effect on the rail transit network, followed by node-degree-first and importance-first recovery strategies, while the random recovery strategy exhibits the worst repair effect. In the human-made damage scenario, the rail transit network is most severely affected, and at this time, the resilience-maximization recovery strategy improves the network resilience by 6.7%, 7.6%, and 25.1% compared to the other three strategies, respectively. Furthermore, when repairing one station, the rail transit network requires approximately 8 hours to recover to optimal performance, while repairing four stations simultaneously only requires about 3 hours to reach optimal recovery. The greater the number of stations that can be repaired simultaneously at the same time point, the faster the network resilience is restored. Implementing the optimal recovery strategy as soon as possible and increasing the repair resource investment in the early stage of the disturbance can significantly improve the recovery efficiency and network resilience of the rail transit network.

Keywords: rail transit; network resilience; recovery strategy; genetic algorithm; adaptive large neighborhood search algorithm