基于最小二乘支持向量回归模型的拉曼光纤放大器优化设计 = Optimization for Raman fiber amplifiers based on least squares support vector regression model

Abstract

在现有拉曼光纤放大器优化设计中，对拉曼散射耦合模型的求解是制约各种优化算法求解效率的瓶颈。提出运用最小二乘支持向量回归机，建立多抽运拉曼光纤放大器的输入输出回归模型，并利用该模型直接替代对拉曼散射耦合模型的求解，实现了优化速率数量级的提升。采用该方法对C 波段的多波长反向抽运拉曼光纤放大器进行了优化实验，仅使用2 个抽运源就可以迅速得到增益波动小于±0.5 dB 的优化设计，并且其平均开关增益可以直接使用该模型进行快速调节。实验结果表明，相比传统的打靶法和平均功率法，通过该拉曼光纤放大器回归模型可以直接得到输出增益，极大地提高了设计的时效性，具有一定的参考价值。 Solving the coupled Raman amplification is the main limitation to the efficiency of the existing optimization method for Raman fiber amplifier design. The least squares support vector regression model is applied to address this issue. Instead of directly calculating the Raman coupled equation, a multi-input multi-output model for Raman fiber amplifier is established. A C-band Raman fiber amplifier with a gain ripple level of ± 0.5 dB is designed using only two pumps. Additionally, the on-off gain can be quickly and directly adjusted by the proposed model. Compared with the shooting algorithm and average power analysis technique, the results indicate that the proposed regression model greatly improves the solving efficiency.