Under the background of the “double carbon” objective and the low carbon transformation of logistics enterprises, a cold chain logistics model based on joint distribution and carbon trading mechanism is constructed. The model takes the multi-depot green vehicle routing problem in cold chain logistics as the research object, and minimizes carbon emission cost, distribution cost and time window penalty cost as the optimization objective. Then, a multiple population genetic algorithm with variable neighborhood search and dynamic catastrophe mechanism is designed to address the disadvantages of poor local search ability and slow convergence of genetic algorithms. And the advantages of the algorithm in terms of optimization ability, stability and convergence speed are confirmed by the standard instances. Finally, an example is analyzed from various perspectives such as joint distribution, objectives and carbon trading mechanism to verify the validity of the model and provide management insights for cold chain logistics enterprises and governments.