Cycle-Stock-Aware Cold-Chain Inventory Management Under Demand Uncertainty

Abstract

Cold-chain inventory management must reduce food loss while maintaining service reliability, product freshness, and food safety for perishable, time-sensitive, and temperature-dependent products under uncertain demand. This study develops a cycle-stock-aware multi-period cost model that separates the average inventory associated with replenishment batches from positive inventory carried across periods. Its objective function jointly evaluates procurement, within-period average-inventory, end-of-period holding, and stockout costs. We examine three operational approaches: a safety-stock heuristic, an ARIMA-based forecast-then-optimize benchmark, and a scenario-based genetic algorithm (GA). The GA is also assessed against an exact LP/MILP benchmark and four metaheuristics: the stellar oscillation optimizer (SOO), dhole optimization algorithm (DOA), phototropic growth algorithm (PGA), and octopus optimization algorithm (OOA). The empirical analysis uses 53 weeks of order data from a group-meal cold-chain enterprise, covering frozen, dry, and chilled product categories. The safety-stock and ARIMA approaches yielded descriptive annual costs of RMB 45.111 million and RMB 46.149 million, respectively; because they follow different evaluation assumptions, these totals are not directly rank-comparable with the scenario-based results. Across 30 runs, the operational warm-start GA averaged RMB 44.684 million, with a coefficient of variation of 0.0146%, and was 0.093% above the exact LP/MILP independent-test cost of RMB 44.643 million. Under a neutral, no-warm-start budget of 8,000 objective evaluations per product, DOA achieved the lowest metaheuristic cost of RMB 45.913 million, 2.845% above the exact benchmark. Training on mean demand rather than demand scenarios increased the independent-test cost by RMB 3.344 million. The study therefore provides a cycle-stock-aware planning and cost-accounting framework while explicitly delineating its empirical scope and limitations.