A non-dominated sorting genetic algorithm for the location and districting planning of earthquake shelters

The improvement of emergency coping capacity is one of the most efficient measures for mitigating disaster impact. Shelter planning is an important strategy to reduce the number of casualties and injuries and facilitate disaster recovery. This study aims to address earthquake shelter location selection and the districting planning of service areas jointly. A bi-objective model is proposed to minimise the total evacuation distance and the total cost, subject to capacity and contiguity constraints. A non-dominated sorting genetic algorithm is developed to tackle the bi-objective model, which involves a multitude of decision variables. To fit the model, the chromosome structure, initialisation process and genetic operators in the algorithm are specifically designed to maintain the contiguity of the service area. And a hybrid strategy of bidirectional multi-point crossover and bidirectional single-point crossover helps promote the diversity of the solutions and accelerate the convergence. Moreover, the Pareto-optimal strategy and feasibility-based rule are combined to obtain trade-offs between objectives. The model and algorithm are validated in a case study of the earthquake shelter location and districting planning problem in Chaoyang District of Beijing, and the results confirm the effectiveness and efficiency of the method.