The main problem to be solved in this article is how to choose a mixed fleet of single and multi-compartment vehicles when delivering multiple orders, and when considering the time-dependence, time window, and deterioration loss, to minimize the total distribution cost of the distribution center, namely the time dependence of multi-compartment vehicle selection problem TDMCVRPTWs. By establishing the heuristic algorithm pre self-organized clustering improved genetic algorithm (SOMGA), and based on the framework of SOMGA algorithm, We build a pre self-organizing clustering improved genetic algorithm of vehicle type selection (SOMGAs), SOMGAs to solve the hybrid fleet that can make the distribution cost the lowest. In order to verify the rationality of SOMGA algorithm, this paper will check the calculation through Solomon’s 25, 50 and 100 customer sets with time windows, and compare the results with HPSO improved particle swarm optimization algorithm. In addition, the optimal values of ACS, HAC, HVANS and HABC are compared. Through analysis, SOMGA algorithm has strong algorithm advantages. In order to verify the feasibility of the pre-self-organization cluster genetic algorithm of vehicle type selection (SOMGAs) based on the framework of SOMGA algorithm, through the Solomon 50 customer, 100 customer set, the hybrid fleet can significantly reduce the distribution cost. And in the process of vehicle class selection, the vehicle path of each vehicle is optimized. Finally, it is concluded that SOMGAs algorithm can solve a reasonable mixed fleet in 80% of cases, reducing the total distribution cost by 0% to 24%. • A new algorithm for solving the multi-compartment vehicle routing problem is proposed. • It provides a new method for studying hybrid fleet. • Various traffic environmental factors and food environmental factors are considered.