Key issues facing enterprises engaged in the supply of fresh agricultural products include the timeliness of the distribution of fresh food, the high cost of this distribution, and predominantly manual scheduling of distribution vehicles. A method to optimize vehicle routes, using the K-means clustering algorithm, was used to develop a model for optimizing transport routes by projecting the urban distribution scenario of fresh agricultural products. The K-means clustering algorithm, an improved genetic algorithm, was introduced into a model for optimizing the distribution paths of fresh produce, which enabled the division of distribution units to be matched with the distribution locations. Consequently, distribution distances and cargo losses were minimized, and the time window as an objective function was not violated. Actual data derived from the distribution orders of a fresh food supply company in Beijing were used in the study. The results of the analysis indicated that without clustering the distribution destinations, the distribution mileage amounted to 3,753.01 km, and the number of vehicles used was 32. The corresponding delivery mileage covered when the delivery destinations were clustered was 2,105.4 km, and the number of vehicles used was 34. When the number of vehicles used exceeded a small number, the total delivery mileage calculated by clustering grouping model was 43.9% lower than that calculated without clustering grouping model. Therefore, it can be concluded that the clustering algorithm based on K-means is suitable for developing distribution scenarios entailing a wide geographical range. A service system for the distribution of urban fresh agricultural products can be designed and developed using the above model. It provides effective means for fresh supply enterprises to reduce distribution cost and improve enterprise efficiency.