基于进食粒子群和共轭梯度的混合优化策略

Mixed optimization strategy based on eating particle swarm and conjugate gradient

传统粒子群算法初期搜索过程中,种群过快地向当前最优粒子飞行,易导致早熟收敛;而算法后期,粒子大量聚集,算法收敛速度慢。通过引入种群进食和二次飞行,提出一种全局性的进食粒子群算法(EPSO),使局部最优附近的粒子进食后快速飞离,以改善种群多样性。并将共轭梯度法(CG)与EPSO相结合形成一种混合优化策略,其中CG用于EPSO的局部搜索过程,以提高收敛速度和精度。利用高维标准测试函数进行寻优实验,并与近年文献方法进行对比,实验结果表明该算法能够克服局部最优的不足,同时继承了CG局部寻优精度高和收敛速度快的特点。

Particle Swarm Optimization （PSO） is an intelligent evolutionary approach widely used to search for the global optimal solution. However, fast flying of swarm particles to the current optimal solution at the early algorithm phase may result in premature convergence, and at the late phase, convergence of a majority of particles causes the degradation of swarm speed. To deal with those shortcomings, a new global algorithm named Eating Particle Swarm Optimization （EPSO） was put forward. In this algorithm, the concepts of eating process and second flight were introduced to guarantee particles flying quickly away from the current optimal solution, so that individual diversity was enhanced. Then the proposed EPSO was combined with Conjugate Gradient （CG） method to form a mixed optimization strategy, in which CG was applied to the local optimization of EPSO algorithm to improve the convergence speed and precision. High-dimensional Benchmark functions were used for optimization experiments, of which the results were compared with the methods in recent literature. The results show that the proposed approach can avoid local optimal phenomena, and obtains the merits of CG in terms of optimization accuracy and convergence speed.