A distributionally robust optimization for blood supply network considering disasters

We study blood supply network optimization considering disasters where only a small number of historical observations exist. A two-stage distributionally robust optimization (DRO) model is proposed, in which uncertain distributions of blood demand are described by a moment-based ambiguous set, to optimize blood inventory prepositioning and relief activities together. To solve this intractable DRO with integer recourse, an approximate way is developed to transform it into a semidefinite program. A case study, based on the Longmenshan Fault in China, validates that our approach outperforms typical benchmarks, including deterministic, stochastic and robust programming. Sensitivity analysis provides helpful managerial insights.