The popularity of fresh e-commerce and cold chain home delivery makes the cold chain logistics orders present the characteristics of "small batch, multi batch, perishable", which further increases the necessity and difficulty of urban cold chain logistics distribution path optimization. Considering both customer satisfaction and road congestion, this paper constructs a mathematical model of cold chain vehicle routing optimization to minimize the total cost. In order to solve this problem, a new knowledge-based ant colony algorithm is designed by integrating the tabu search operator under the knowledge-based elitist strategy and the knowledge model of dynamic probability selection into the ant colony algorithm. Through the simulation experiments of simulation examples and real examples, the traditional ant colony algorithm, the improved ant colony algorithm based on tabu search and the knowledge based ant colony algorithm proposed in this paper are compared and analyzed, and the effectiveness of the proposed model and knowledge-based ant colony algorithm is verified.