To solve the problem of huge distribution demand and “last mile” distribution, this paper first proposes the cooperative delivery of multiple vehicles and multiple drones with dynamic energy consumption(CDMVMD-DEC), and provides a mixed integer programming model(MIP) aimed at minimizing the delivery time. To solve the problem efficiently, the adaptive large neighborhood search(ALNS) based on problem domain knowledge is proposed, along with the combination of the K-means clustering and the nearest neighbor for constructing the initial solution. Experiments on different-scale instances demonstrate that the ALNS outperforms the simulated annealing, variable neighborhood search and genetic algorithm in solution quality and computational time. In terms of solution quality, the performance of the ALNS is improved by 23.8$%$, 23.3$%$ and 5.7$%$ respectively. The results of experiments show that ALNS provides a better balance between global search and local search. Moreover, the results of the sensitivity test show that the load capacity and endurance of drone are the important factors affecting the delivery time.