The work aims to ensure the customer satisfaction of fresh delivery while reducing the distribution costs. Based on the comprehensive consideration of such factors as transportation distance, distribution time window constraints and deterioration of fresh food, with transportation costs, refrigeration costs, vehicle call costs, distribution center opening and idle costs, and the satisfaction of self pick-up site in terms of distribution time and fresh quality as functions, a dual-objective fresh delivery route optimization model was established. The non-dominated sorting genetic algorithm was used to obtain the fresh delivery plan under different demand conditions. Under different demand conditions, reasonable selection of open distribution centers can effectively reduce distribution costs while ensuring certain customer satisfaction.