差分演化算法在约束优化问题中的应用

差分演化算法的变异机制没有充分利用种群的信息,导致变异是盲目的。受到粒子群算法信息共享机制的启发,文中提出了一种多群体差分演化算法,新算法将整个种群分成多个子种群,每个子种群通过借鉴本种群的内部经验与整个种群的外部经验对变异进行指导。一方面,由于变异操作借鉴了子种群的局部信息和整个种群的全局信息,提高了算法收敛的速度;另一方面,多个子群体增强了种群的多样性,提升了算法的全局搜索能力。数值实验表明新算法具有很强的稳定性和全局搜索能力,在相同计算复杂度情况下的全局搜索能力较原始差分演化算法有明显提升,可以有效求解约束优化问题。

Application of Differential Evolution Algorithms on Constraint Optimization Problems

The mutation mechanism of differential evolution algorithm doesn't apply the population information sufficiently,so the mutation operation is blind.Inspired by particle swarm optimization's information sharing mechanism,propose a multiple population differential evolution algorithm,which divides whole group into many sub-populations,thus the every sub-population's experiences from referring on inner message and outer message could instruct mutation operation.On the one hand,accelerating the convergence;on t...