The blood supply chain is recognized as one of the most complex healthcare networks due to high perishability, time sensitivity, reliance on voluntary donation, and demand fluctuations, making it increasingly vulnerable in critical situations such as natural disasters or pandemics. This study aims to enhance the sustainability and resilience of this network by proposing a hybrid framework based on fuzzy demand forecasting and robust multi-objective optimization. In the first step, a Long Short-Term Memory (LSTM) neural network is employed using real data from five blood transfusion centers to forecast daily demand in the form of three fuzzy scenarios—optimistic, baseline, and pessimistic—achieving a mean absolute percentage error (MAPE) of approximately 4.42%. These outputs are then incorporated into the optimization model. The developed mathematical model simultaneously optimizes location decisions for permanent and temporary centers, donor allocation, and blood flow management, with three main objectives: minimizing total cost, reducing blood shortages, and minimizing environmental pollution. The NSGA-II metaheuristic algorithm is utilized to extract the Pareto front, while CPLEX and MOPSO are employed for model validation on small-scale instances. The proposed framework successfully generates a Pareto front comprising seven non-dominated solutions, achieving favorable convergence and uniform diversity. Convergence analysis based on ten independent runs indicates full stabilization of objective functions after 200 generations, confirming Pareto front stability. Sensitivity analyses and evaluation of sustainability and resilience indices demonstrate that the proposed model provides adaptive and stable performance under critical conditions through network reconfiguration. The model is generalizable to other regions and can be effectively utilized in strategic healthcare planning.