Abstract: Particle swarm optimization algorithm based on variable metric method is proposed for the defects of elementary particle swarm optimization algorithm "premature" and the parameter setting.The algorithm uses fast local convergence characteristics of the variable metric method，so that the improved algorithm can jump out of local optimal solution effectively，and can also search the global optimal solution quickly.Simulation results show that the new algorithm improves the accuracy of the optimal solution and optimization efficiency;also demonstrate that the new algorithm has better robustness，and then the improved algorithm is successfully applied to solve the problem of nonlinear equations.

Key words: global optimization, particle swarm optimization algorithm, variable metric algorithm, nonlinear equations

摘要： 针对基本粒子群优化算法的“早熟”及参数设置的缺陷，提出基于变尺度的粒子群优化算法。该算法利用变尺度法局部收敛快的特点，使改进后的算法能有效地跳出局部最优解，快速地搜索到全局最优解。仿真结果表明新算法提高了最优解的精度和优化效率;同时验证了新算法有较好的鲁棒性，然后把改进算法成功应用于非线性方程组求解问题。

关键词: 全局优化, 粒子群优化算法, 变尺度算法, 非线性方程组