基于随机p-鲁棒优化的国家血液战略储备库选址-库存模型

以往的救灾实践对建立国家血液战略储备体系提出了迫切要求。国家血液战略储备库的建设问题亟待解决。由于血液产品特性以及应急血液保障特性的存在，使得国家血液战略储备库的选址决策具有一定的复杂性。本文将问题定位为选址-库存问题。首先，以应急条件下血液保障及时度最高为目标，构建了一个不确定环境下考虑多情景、多血型、多阶段、带提前期、有容量限制、日常随机需求、有预算约束及协同定位的国家血液战略储备库选址-库存模型。同时，为了规避应急条件下的不确定风险，进一步构建了国家血液战略储备库选址-库存问题的随机p-鲁棒优化模型。该模型为离散非线性混合整数规划模型，难以快速精确求解。故基于模型性质，设计了相应的遗传算法。最后，设计了两组算例验证模型与算法的有效性。其中，第1组算例基于我国大陆地区31个省级血液中心与省级行政区的数据，并根据不同预算值给出6个算例，得到了国家血液战略储备库的选址-库存决策方案。第2组算例为6个不同规模的模拟算例，用来测试不同规模下的算法性能。算例结果表明：遗传算法的性能更好；鲁棒解与确定性模型最优值相差不大（最大差距≤1.08%），可降低不确定性导致的风险。实践中，可对本文所建模型稍作改进，应用于具有类似特征的易腐品（药品、粮食等）应急物资储备库选址-库存决策。

Location-inventory Model of National Blood Strategic Reserves based on p-robust Stochastic Optimization Method

The practice of emergency rescue has put forward impending requirements for establishing the national blood strategic reserves (NBSR). An emergency blood support system urgently needs to be constructed in China. However, the characteristics of the blood products and emergency blood support make the location decision-making of NBSR complicated. Hence, a joint LIP is constructed by considering the lead-time, multi-scenarios, multi-stages, multi-blood types, daily stochastic demands, facility capacity constraints and coordinated location with aiming at maximizing timeliness of blood supply under unconventional emergencies. Then, a genetic algorithm is further developed based on the discrete nonlinear mixed integer programming model mentioned above. Finally, two sets of numeral examples is conducted to verify the effectiveness of the model and algorithm. The first set of numeral examples are conducted based on the data which from 31 provincial-level blood centers and provincial-level administrative regions in Chinese Mainland. And 6 numeral examples are given according to different budget values to initially make a location-inventory decision. The second set of numeral examples contains 6 simulation studies under different scales, which are used to test the performance of algorithm proposed in this study. The result shows that the genetic algorithm designed above has a better performance, which can lead to a very small gap (≤ 1.08%) between robust solutions and optimal values of the deterministic model and therobust optimization can reduce the uncertainty risk. In practice, the model constructed in this study can be revised and improved to provide a decision support in solving location-inventory decision problem about relief supplies reserve bases of perishable goods (such as drug, food, etc.) with similar characteristics.