多种群随机差分粒子群优化算法及其应用

Multi-population random differential particle swarm optimization and its application

为提高粒子群算法的寻优性能,提出了一种新的多种群随机差分粒子群优化算法。该方法将种群随机分组,利用基于吸引概率的轮盘赌方法确定其可能搜索方向。寻优效果预期不明显时,进行子种群内部随机差分进化寻优,以增加寻优方向的随机性和多样性。并给出了一种新的约束处理方法,对种群粒子进行动态划分,仅对部分粒子进行速度更新和位置更新,提高了搜索速度。并将所提出算法应用于数值优化问题和焊接梁设计问题。仿真结果表明,所提出算法在处理多峰函数问题时,寻优精度高,收敛速度快。在处理有约束问题时,提出的处理约束的方法,明显缩短了寻优时间。算法在处理复杂的无约束问题和有约束问题上均具有很好地寻优性能。

To improve the performance of particle swarm optimization, a new multi-population random particle swarm optimization,based on differential evolution,is proposed in this paper. The proposed algorithm randomly di-vides the population into several groups and uses the roulette wheel method,based on attraction probability,to de-termine the possible search direction. When the expected optimization effect is not obvious,this algorithm uses ran-dom differential evolution to generate a new solution in the sub-population,aiming to increase the randomness and diversity of the search direction. In addition,a new constraint handling method is proposed to dynamically divide populations. Particle implemented velocity and position updates are proposed to improve search speed. Finally, the proposed algorithm is applied to a numerical optimization problem and a welded beam design problem. Simulation results show that this algorithm has the advantage of high precision and fast convergence when dealing with multi-peak functions. For dealing with a constrained problem,a new method is proposed to handle the constraints,which markedly shortens the search time. The proposed algorithm shows good optimization performance for complex con-strained and unconstrained problems.