Research And Application Of Parallel Artificial Bee Colony Algorithm

Swarm intelligence optimization algorithm is one kind of random search algorithms,which can solve the optimization problem that can not be solved by traditional optimization technologies.Therefore,many experts and scholars show their favor on it.Artificial bee colony algorithm is a new swarm intelligence optimization algorithm.The algorithm is mainly to simulate the foraging behavior of bees in the real world to get the optimal solution.As the artificial bee colony algorithm was put forward,its characteristics of few parameters setting,simple calculation,good parallelism,robustness results in a good effect in dealing with optimization problem,thus,many domestic and foreign scholars pay attention to it.Although artificial bee colony algorithm has many advantages in dealing with optimization problems,there are still some problems such as easy to fall into local optimal solution and premature convergence.In particular,when the artificial bee colony algorithm is used to deal with the optimization problems of complex high dimension and large scale,it is hard to accept the long run time.Therefore,in order to solve these problems,many scholars begin to research on parallel algorithm of artificial bee colony algorithm.It is necessary to improve the efficiency and accuracy of optimization algorithm by parallel artificial bee colony algorithm.The parallelization of the artificial bee colony algorithm can be realized by a variety of parallel methods.At present,scholars at home and abroad mainly use the multi-cluster MPI(Messing Passing Interface)technology or use the stand-alone Java multi-threading technology to achieve coarse-grained parallel artificial bee colony algorithm.In the published papers,they pointed out that the main time consumption of artificial bee colony algorithm is in the calculation of the fitness function in dealing with high dimension optimization problem.This is the focus and difficulty of studying the parallelization of artificial bee colony algorithm.After discussing the parallelization techniques,the PCABC(Parallel Changed-in-selection-mode Artificial Bee Colony)algorithm is proposed to solve this problem by using OpenMP(Open Multi-Processing)multi-threading technology,and combining the improved mechism of onlooker bees choosing employeed bees.By using the OpenMP method of master and slave parallel model and shared memory,and the time-consuming part of the computational fitness function were made to parallel execution,which can greatly reduce the processing time in dealing with high-dimensional optimization problem.The experiment was conducted in three different kinds of task scheduling methods,respectively,considering that the parallel computing of the fitness function corresponding to three different kinds of task scheduling methods.The experimental results show that the parallel transformation of calculating fitness function of artificial bee colony algorithm can greatly speed up the processing time and greatly accelerate the convergence speed in dealing with high dimension problems.The experiment has achieved the anticipated target.After proving the effectiveness of the parallel artificial bee colony algorithm,the PCABC algorithm is applied to solve the problem of high dimensional function optimization and watershed hydrological model optimization.The parameter optimization rate has a crucial effect on the overall performance and hydrological forecast results of the hydrological model.There are a lot of computationally intensive tasks in the optimization of model parameters,which need to consume a lot of CPU processing time.Therefore,the PCABC algorithm was applied to the Xin'anjiang two-source model to optimize the hydrological model.The experimental results show that PCABC algorithm can significantly improve the efficiency and accuracy of the parameters of the hydrological model,and has the advantages of low parallel cost and simple implementation process.PCABC algorithm has a good performance in parameter optimization of hydrological model.It can be an effective and feasible method to solve the parameter optimization problem of hydrological model.