学习因子随权重调整的混合粒子群算法

Research of Simplified Particle Swarm Optimization Algorithm with Weight and Learning Factor

首先使惯性权重随迭代次数和粒子状态非线性改变平衡算法的全局探测和局部开采的能力,为了解决惯性权重与学习因子独立调整削弱了粒子群算法的统一性和智能性等问题,通过分析惯性权重与学习因子的变化关系,将学习因子表示为惯性权重的logistic回归分析型函数。由于非线性因子的加入会降低粒子的多样性,结合差分进化算法的交叉算子和变异策略,利用交叉算子来提高算法的全局探索能力,保持种群多样性;利用差分进化算法的变异策略产生候选解来更新位置公式,给出了学习因子随权重调整的混合粒子群算法,并对新提出算法的收敛性进行理论分析。将此改进算法与相关算法在四个测试函数上进行对比实验,证明该算法在寻优精度、迭代速度和收敛成功率上有明显改进。

Firstly,the inertia weight changes the global detection and local mining capability of the equilibrium algorithm nonlinearly with iteration times and particle state.In order to solve the problem that the independent adjustment of inertia weight and learning factor weakens the unity and intelligence of particle swarm optimization(PSO)algorithm,the learning factor is expressed as a logistic regression analysis function of inertia weight by analyzing the relationship between inertia weight and learning factor.Since the addition of nonlinear factors will reduce the diversity of particles,combining the crossover operator and mutation strategy of the differential evolution algorithm,the crossover operator is used to improve the global exploration ability of the algorithm,which keeps the diversity of the population.By the variation strategy of the differential evolutionary algorithm,the candidate solutions can be generated to update the position formula.A hybrid particle swarm optimization algorithm with learning factors adjusted with weights is proposed and its convergence is analyzed theoretically.Finally,the improved algorithm is compared with the existing algorithm on four test functions,and it is proved that it has obvious improvement on the optimization accuracy,iteration speed and convergence success rate.