Improved Artificial Bee Colony Algorithm Based On Search Strategy And Its Application

Artificial bee colony algorithm(ABC)was first proposed by Karabogo in 2005 due to the influence of bee colony behavior.As a new swarm intelligence optimization algorithm,ABC algorithm has attracted the attention of many scholars due to its advantages such as fewer control parameters,strong exploration ability and easy implementation.However,ABC algorithm also has some problems,such as poor local search performance,precocious convergence and low convergence accuracy.Therefore,the research on basic artificial bee colony algorithm still has broad prospects.For unconstrained optimization problems,steepest descent method,conjugate gradient method and simplex method are respectively used to enhance the local search ability of basic ABC algorithm,then four improved ABC algorithm are proposed.In the scout bee stage of the basic ABC algorithm,the employed bees change role into scout bees,the food sources that have reached the limits of exploitation have been updated by random search of scout bees,that can not certain that the updated food source is better,but updating the food source with the steepest descent method and the simplex method can ensure that all updated food sources are improved,the improved artificial bee colony algorithm with steepest descent method and the improved artificial bee colony algorithm with simplex method based on gaussian perturbation are proposed.In basic ABC algorithm,the onlooker bees will choose and update the food source according to the feedback from the employed bees,the random search method adopted by the onlooker bees for the food source cannot guarantee that the food source location after the search is better than before.However,conjugate gradient method can ensure that each iteration is moving in a better direction,therefore,the location of food source obtained by using conjugate gradient method is bound to be better,an artificial bee colony algorithm based on conjugate gradient method is proposed.In the onlooker bee stage of the basic ABC stage,the random search method adopted may make the updated food source location worse,therefore,the simplex method is introduced to the ABC algorithm,and the current best food source location is taken as a guide to make the updated food source approach to the best location.At the same time,the optimal strategy was added in the scout bee stage,that is,the food source that have reached the limits of exploitation is first determined to be the current best food source location.If it is,it is reserved.If not,it is searched and updated by scout bees,an improved artificial bee colony algorithm based on simplex method of optimal individual guidance is proposed.All the four algorithms replace random search with deterministic search,which strengthens the local search ability of the basic ABC algorithm and avoids the algorithm falling into the local optimum.The standard test problem is used to carry out numerical experiments on the four improved algorithms,and the results show that the four improved algorithms have higher accuracy and better stability than the basic ABC algorithm and some other swarm intelligence optimization algorithms.For constrained optimization problem,this paper proposes an improved ABC algorithm(PF_ABC)based on penalty function.In this algorithm,penalty function is introduced,adaptive dynamic penalty factor is set,and the value of objective function and constraint violation degree are normalized,finally,a new objective function is constructed by using the objective function and constraint violation degree.The constrained optimization problem is transformed into multi-objective function optimization problem.The PF_ABC algorithm is numerically tested by using the standard test problem.The results show that the PF_ABC algorithm is of high accuracy and stable performance.