Aiming at the problems of easy falling into local extremum and low convergence speed,we propose a self-adaptive step flower pollination algorithm (FPA) based on the simplex method.We use the adaptive step length strategy to update individual location in the global optimization of the FPA,and the step is decreased dynamically along with the increase of iteration so that it can effectively avoid local optimum.In the process of local optimization of the FPA,to enhance the capacity of global optimization and improve the convergence speed,we perform expansion and contraction/compression operation of the simplex method on the weak individuals that enter the next iteration.The comparison and analysis of simulation results on the 8 CEC2005 benchmark functions show that the proposed algorithm has better global searching ability and robustness,and faster and more precise convergence than those of the basic flower pollination algorithms.