Distributed PMD Measurement Based on Machine Learning Assisted POTDR

A method for distributed polarization mode dispersion (PMD) measurement based on machine learning assisted POTDR is presented in this study, which extracts the characteristics of polarization optical time-domain reflectometer (POTDR) curves and inputs them into a pre-trained machine learning model to obtain distributed PMD values. Compared with traditional distributed PMD measurement methods, this proposed method achieves an average local differential group delay (DGD) measurement accuracy of 0.05 ps/km over 50 m, while improving measurement efficiency by 75% compared to methods with similar accuracy, and significantly enhancing the simplicity and efficiency of distributed PMD measurement. The performance of multiple machine learning models for distributed PMD measurement is compared, providing a basis for selecting models for real-time measurement of distributed PMD in optical fiber composite overhead ground wire (OPGW) under different dynamic environments.