光学学报, 2020, 40 (23): 2314002, 网络出版: 2020-11-24
基于遗传退火算法的光纤激光主动偏振控制技术研究 下载: 1015次
Research on Active Polarization Control System of Fiber Laser Based on a Mixed Genetic Algorithm and Simulated Annealing Algorithm
摘要
介绍了一种用于偏振控制的遗传退火算法(GASA),该算法在遗传算法(GA)的基础上,引入了模拟退火的选择机制,将蒙特卡罗思想引入到了GA中,使算法具有更加强大的搜索能力。通过建立基于GASA算法的光纤激光主动偏振控制系统的数学模型,得到了不同参量情况下的GASA算法的仿真图像,并且分析了其收敛效果。仿真结果表明,在选择输出激光的偏振消光比作为适应度函数,种群数量为90,变异概率为0.7,交叉概率为0.001,温度下降比率为0.99的情况下,系统可以达到最优的控制效果。将GASA算法和随机并行梯度下降算法的仿真图像进行对比,可以看出,GASA算法具有较好的全局搜索能力和跳出局部最优值的能力,可以将其用在光纤激光的主动偏振控制系统中。
Abstract
Therefore, this paper proposed a mixed genetic algorithm and simulated annealing (GASA) algorithm for polarization control. Based on a genetic algorithm (GA), this algorithm introduced the selection mechanism of simulated annealing (SA), and introduces the Monte Carlo idea into GA, which made the algorithm have more powerful search ability. By establishing a mathematical model for the active polarization control system of fiber laser based on the GASA algorithm, we obtained the simulation images of the GASA algorithm under different parameter combinations and analyzed the convergence effect. The simulation results show that when the polarization extinction ratio of output laser is selected as the fitness function, the population number is 90, the mutation probability is 0.7, the crossover probability is 0.001, and the temperature drop ratio is 0.99, the system can achieve the optimal control effect. By comparing the simulation images of GASA algorithm and stochastic parallel gradient descent algorithm, we can see that the GASA algorithm has better ability of global search and jumping out of local optimal values, and can be used in the active polarization control system of fiber laser.
尤阳, 漆云凤, 沈辉, 邹星星, 何兵, 周军. 基于遗传退火算法的光纤激光主动偏振控制技术研究[J]. 光学学报, 2020, 40(23): 2314002. Yang You, Yunfeng Qi, Hui Shen, Xingxing Zou, Bing He, Ju Zhou. Research on Active Polarization Control System of Fiber Laser Based on a Mixed Genetic Algorithm and Simulated Annealing Algorithm[J]. Acta Optica Sinica, 2020, 40(23): 2314002.