Abstract: This paper makes an analysis about the defects of particle swarm optimization algorithm easily relapsing into local optimization and convergence with low speed as well as low precision in the late evolution period.Based on differences of global research and local research capacity made by the inertia weight of diverse particles in the same iteration period，a method that individual inertia weights adjust with the individuals’ fitness is put forward.What’s more，in the late evolution period，in order to prevent the population from becoming bad and easily relapsing into local optimization，Cauchy mutation operator to escape local fitness is brought forward，and then，according to the test and comparison with four typical functions，this method is illustrated with higher accuracy and faster convergence rate.In the end，this modified method is applied to wind farm wind speed probability distribution modeling，and compared with traditional statistic strategy，an illustration with higher precision is given.

摘要： 分析了微粒群算法易陷入局部最优和进化后期收敛速度慢且精度较低的缺陷，针对即使在同一迭代时期，不同的微粒的惯性权值所调节的局部搜索能力和全局搜索能力的差异的基础上，提出了一种在同一迭代时期，个体惯性权重随个体的适应值自适应调整的方法。另外，在进化后期，防止种群多样性变差和易陷入局部最优，提出了柯西变异算子，使这些微粒能及时跳出局部极值，通过典型的4种测试函数的测试和比较，表明改进的算法精度更高，收敛速度更快。然后将改进的微粒群算法用于风电场风速概率分布模型的优化，与常规的统计方法相比，表明该方法具有更高的拟合精度。