The optimization problem exists extensively in various scientific fields.To solve these problems,various optimization algorithms have been proposed.Flower Pollination Algorithm(FPA)is an optimization algorithm that realizes swarm intelligence by simulating natural plant flower pollination process.The algorithm has many advantages such as fewer parameters,simple structure and easy implementation.FPA has attracted the attention of researchers and has been applied in various fields.However,the FPA has the problems of slow convergence,low local search efficiency,and easy to fall into local optimal.This paper analyzes the causes of the insufficiency of the FPA algorithm and improves on these deficiencies.In order to enhance the FPA's optimization ability,this paper improve FPA from five aspects according to the structure of the algorithm: First,in the population initialization stage,the population is divided into three equal parts,the first part of the individuals is randomly generated,the second part of the individuals is uniformly and randomly generated,in the three parts,the optimal individuals in the first two parts are selected to perform elite reverse initialization.Second,in the conversion probability,the population diversity is calculated based on the distances from all individuals in the population to the optimal individual,and the conversion probability is calculated according to population diversity.Third,in the global optimization section,the algorithm use the beetle antennae search to accelerate the convergence speed of FPA;Four,in the local optimization section,the original differential strategy is changed to another differential strategy that the global optimum value and the current individual participate in.In addition,a small-probability mutation strategy was introduced to randomly mutate individuals without affecting the global optimum.This improved the population diversity and helped the algorithm to jump out of the local optimum.Five,in the cross-border processing,the individual's cross-border dimensions were mutated to prevent individuals from gathering at the border.In conclusion,this paper proposed a flower pollination algorithm based on beetle antennae search and mutation strategy(BMFPA).In order to verify the effectiveness of the improvements of the algorithm,a number of test functions were used to test the BMFPA with high-dimension,low-dimension and fixed precision.The results show that BMFPA has an accuracy improvement of 7 to 20 orders of magnitude in lowdimensional than the original FPA.BMFPA has an accuracy improvement of 6 to 11 orders of magnitude in high-dimensional than the original FPA.The number of iterations required to achieve the target accuracy is less than the original FPA.In order to verify the effectiveness of the improvements,the experiment used the control variable method to perform the hybrid population initialization,the adaptive transition probability,the BAS improving in the global optimization,the mutation strategy improving in the local optimization and the boundary mutation.Comparing with the original algorithm,the experimental results show that all the improvements have the effect of improving the search accuracy.Among all,BAS and mutation strategy have the greatest improvement in FPA algorithm optimization accuracy. |