A Hybrid Convolutional Neural Network and Particle Swarm Optimization Framework for DFB Laser Parameter Extraction

Smart grid applications are increasingly reliant on optical sensing and communication systems for real-time monitoring and control. The stability and accuracy of these systems are profoundly influenced by the performance of distributed feedback (DFB) lasers, which serve as their optical sources. However, the nonlinear coupling between the optical and carrier fields makes parameter extraction for DFB lasers a challenging task. This paper proposes a hybrid framework integrating convolutional neural networks and particle swarm optimization (CNN-PSO) for efficient parameter extraction of DFB lasers. The method combines the global search capability of PSO with the nonlinear representation ability of CNNs, enabling accurate extraction of key physical parameters. The papers provide a practical foundation for enhancing the stability and calibration accuracy of optical modules in smart grid sensing and communication subsystems.