Agri-food supply chains are increasingly vulnerable to disruptions from climate change, geopolitical conflicts, and pandemics, which manifest as both supply-side (e.g., volatile lead times) and demand-side uncertainties, threatening profitability and food security. While existing research has addressed these risks in isolation, a critical gap exists in tactically configuring perishable food supply chains to jointly mitigate both types of risk through integrated decision-making on sourcing, logistics, and inventory. This paper develops a novel Food Supply Chain Configuration Problem under Supply–Demand Risks (FSCCP-SDR). It is the first optimization model to simultaneously determine operational mode selection (supplier, transport, processing) and multi-echelon inventory placement (safety stock and early arrival stock) for perishable products under both stochastic demand and stochastic lead time, maximizing total net profit. To solve this NP-hard problem, we propose an exact dynamic programming method for small instances and a hybrid Particle Swarm Optimization-Genetic Algorithm (PSO-GA) with a custom repair mechanism for scalability. Applied to a grape processor case study in China, the hybrid PSO-GA finds near-optimal solutions (within 0.57% of optimal) efficiently. The model yields non-intuitive, data-driven configurations, such as selecting low-cost, high-loss options upstream and prioritizing reliability downstream. Sensitivity analyses reveal: (1) increased supply-side risk elevates both safety and early arrival stock, justifying investments in supplier reliability; (2) extended shelf-time significantly reduces required inventory, quantifying the value of preservation technologies; and (3) under extreme perishability, the optimal strategy pivots from "buffering with inventory" to "buffering with responsiveness." The methodology demonstrates strong scalability on larger synthetic networks.
Nie et al. (Wed,) studied this question.