In order to solve the problem that the traditional cold chain VRP defaults the distribution area to a two-dimensional plane, which makes the actual scheduling different from the ideal state and causes the distribution cost to increase, the work aims to make an in-depth research on related problems, aiming at rationally planning the allocation of resources and improving the economic benefits of enterprises. The fuel consumption cost and carbon emission cost were calculated by introducing the slope function, and various realistic constraints such as three-dimensional space location, multiple models, load changes, customer's pickup and delivery requirements, and time window requirements were comprehensively considered. The improved genetic algorithm of K-means multi-dimensional spatiotemporal clustering was solved. The solution results indicated that if the road slope was initially considered in the cost calculation, the distribution cost could be effectively reduced by 9.58%. At the same time, large vehicles should avoid sudden altitude changes when heavily loaded. When the altitude changed, large vehicles could change to a higher altitude after some goods were unloaded. The model studied is closer to the actual road network and has higher applicability, especially for the distribution of cold chain products in cities with obvious road slopes, which has practical reference significance.