融合 Rosenbrock 搜索法的混合粒子群算法

为克服粒子群算法在处理复杂高维问题时易陷入局部最优及寻优精度低等缺陷，提出一种融合 Rosenbrock 搜索法的混合粒子群算法。首先，利用 Tent 混沌序列进行种群初始化；其次，采用去速度项的简化粒子群公式提高收敛速度并对个体极值加入扰动，增强粒子种群多样性；最后，当全局最优个体更新停滞时，利用Rosenbrock 搜索法对全局最优个体进行局部搜索，提高解的精度。利用 8 个常用基准测试函数分别对 30 维和50 维问题进行实验，证实该算法可寻到病态函数 Rosenbrock 全局最优值，且比其它 7 个函数的寻优精度提高10-2 数量级。实验证明该算法收敛速度快，解的精度高，全局搜索能力强，寻优能力明显提高。

Hybrid Particle Swarm Optimization Algorithm Combining Rosenbrock Search Method

For the particle swarm algorithm，it is easy to fall into local optimum and low precision in processing when dealing with complex and high-dimensional problems. A hybrid particle swarm optimization algorithm combining Rosenbrock search is proposed. Firstly，the improved algorithm uses the Tent chaotic sequence to initialize the population，so that the initial particles are well distributed in the solution space. Secondly，the simplified particle swarm optimization algorithm with the velocity term removed is used and the disturbance is added to the individual extremum to enhance the particle population diversity，so that the algorithm is not easy to premature. Finally，the global optimal individual’s stagnation update times is used to judge whether the algorithm falls into local optimum，and the Rosenbrock search method is used to locally search the global optimal individual to improve the solution quality. Experiments were performed on 30-D and 50-D problems using eight common benchmark functions. Compared with some improved particle swarm optimization algorithms proposed in other literatures，the proposed algorithm can find the global optimal value of the ill-conditioned function Rosenbrock，and the optimization accuracy of the other seven functions has an improve of magnitude of 10-2 . The proposed algorithm has fast convergence speed，high precision of the solution，strong global search ability and obvious improvement ability.