Differential evolution Algorithm has been widely used due to its efficient performance. However, different mutation strategies and parameter settings can greatly influence the performance of DE. In order to weaken the impact of a specific mutation strategy on DE algorithm and to fully leverage the advantages of existing mutation strategies, this article designs a new mechanism, which can adaptively choose mutation strategy by using selection probability memory(SPM) in the process of evolution. By combining this mechanism with adaptive parameters’ setting, a new adaptive DE algorithm is proposed(AMDE). In AMDE, the process of evolution is divided into three stages, various mutation strategies are available for choice at each stage. AMDE improves the performance by combining the advantages of multiple mutation operators. The performance of AMDE is evaluated on 29 functions. Compared to differential evolution variants and other state-of-the-art algorithms, the experimental results show that AMDE is superior.