Low-carbon cold chain distribution facility location and multi-temperature time-dependent routing optimization based on bi-level programming

Abstract To address the high energy consumption and carbon emissions in cold chain logistics, this study develops a bi-level programming model for coordinated optimization of distribution center location and multi-temperature routing. The upper-level model employs an improved adaptive immune algorithm to optimize distribution center location, reducing construction costs by 2.25% and average delivery distance by 6.09% via reverse learning initialization and dynamic crossover mutation probabilities. The lower-level model constructs a time-varying road network based on congestion delay index change rates and utilizes a hybrid ant colony algorithm combined with adaptive large neighborhood search to optimize multi-temperature delivery routes, reducing delivery distance by 14.27 km and distribution costs by 3.55%. Validated with 79 customer points in Zhengzhou, the model identified three optimal distribution centers with a 91.4% vehicle loading rate. Compared to static road networks, total mileage and distribution costs decreased by 3.5% and 6.0%, respectively. Comparative experiments on the Solomon dataset demonstrate the proposed algorithm’s superiority in total travel distance and average path length, with sensitivity analysis further confirming the algorithm’s robustness. The results indicate that this model effectively integrates location decisions with route optimization. The improved algorithm exhibits superior performance in solution accuracy and stability, providing theoretical support and practical guidance for cold chain enterprises to achieve low-carbon and efficient operations.