The differential evolution(DE) algorithm has some disadvantages, such as easy to fall into local optimal solutions, slow convergence speed and low convergence precision in the later stage of evolution. By means of combining DE/rand/1 mutation mode with global exploration ability and DE/best/1 mutation mode with local exploitation ability, and introducing the elite archive strategy and control parameters adaptation strategy, a self-adaptive differential evolution with dual mutation modes collaboration(DMCSaDE) is proposed. A total of 15 typical benchmark test functions are used to perform comparative experiments. The experimental results show that DMCSaDE can effectively improve the global exploration ability and local exploitation ability, avoid the premature convergence and greatly improve the convergence performance and robustness. Additionally, the size of elite population has a significant effect on the optimization performance of DMCSaDE.