关键词: 地震救援, 车辆路径规划, 应急物资调度, 大邻域搜索模拟退火混合算法

Abstract: In recent years, large-scale natural disasters have frequently occurred in China, seriously threatening the lives and property of the people. The field of emergency management has gradually attracted the attention of scholars. The Chinese government increasingly emphasizes concepts such as “people-oriented, life comes first” and “saving lives is the first priority of emergency rescue” in emergency management. However, in the field of post-earthquake rescue, there are still very few studies that directly use the number of deaths as an optimization goal. An efficient emergency dispatch plan can greatly reduce the mortality rate caused by earthquake disasters. However, the disaster situation at each disaster site is different, and there is also a certain difference in the urgency of the demand for rescue resources. Therefore, after an earthquake occurs, how to comprehensively consider the disaster situation at various disaster sites and determine the optimal path plan for the deployment of rescue vehicles to minimize personnel casualties caused by disasters is of practical significance.
This paper takes earthquake events as the research background, builds a mixed integer programming model intending to minimize the number of deaths based on the prediction of the number of casualties and material needs at each disaster site, and analyzes vehicle travel time during the instability of the post-earthquake road network, in the consideration of time-varying disaster conditions, and by the introduction of the “death rate-time” function. An improved hybrid optimization algorithm is designed to solve the model. The introduction of the “death rate-time” function can establish the relationship between “rescue vehicle arrival time” and “cumulative death rate at disaster site”, with “rescue vehicle arrival time” as an intermediate variable to make “cumulative death rate at disaster site” a decision variable and determine the number of deaths when rescue vehicles arrive at the disaster site. This objective function embodies the core concept of “life comes first” in emergency rescue and is also significantly different from the objective functions of most existing VRP variants. The numerical experiments have shown that the model proposed in this paper can effectively reduce the number of deaths compared to traditional rescue time minimization models, with an average reduction rate of 30.85%. This study is a variant of the VRP problem and belongs to the NP-hard problem, so an algorithm needs to be designed to solve it. This paper designs an optimization algorithm that combines large-scale neighborhood search and simulated annealing and improves the destruction operator of the large neighborhood search algorithm and the cooling function of the simulated annealing algorithm. Statistical indicators are used for comparative analysis in numerical experiments. The results show that compared with conventional algorithms, the hybrid optimization algorithm proposed in this paper has optimization rates of 2.8% and 3.5% on optimal value and average value respectively, proving that the designed algorithm has better global search performance.
The numerical experiments integrate the research work of the whole paper and prove that the model and algorithm proposed in this paper could provide decision-makers with scientific and reasonable optimization schemes for post-earthquake emergency material dispatch. Moreover, the research of the whole paper also suggests that post-earthquake emergency rescue decision-making should not have simply considered time-saving, but should have comprehensively considered the disaster situation and rescue time at each disaster site.

Key words: earthquake rescue, vehicle routing problem, emergency material dispatching, large neighborhood search-simulated annealing hybrid algorithm