Raman Tilt Prediction for Digital Twin Modelling of ROADM-based Transmission Systems

Fulfilling the anticipated demands of future network technologies entails a plethora of challenges. Traditional approaches to tackle these challenges struggle with scalability in disaggregated systems, leading to a shift towards data-driven methods, especially machine learning (ML). Digital twins offer a cost-effective way to optimize network management and support decision-making. However, accurately predicting physical phenomena like stimulated Raman scattering (SRS) remains a challenge, as conventional physics-based models may fail due to variations in transmission spans. In this paper, we develop a deep neural network (DNN) model to predict the SRS-induced Raman tilt in optical fibers. The DNN is trained with data obtained from experiments on the COSMOS testbed, and its performance is benchmarked against established analytical models. Our results demonstrate that the proposed DNN has a mean absolute error between 0.03 dB – 0.12 dB.