Digital-twin-enabled launch power optimization in a submarine SDM optical network: field demonstration

Launch power optimization in submarine space-division multiplexing (SDM) networks is hindered by spatial–spectral coupling, nonlinear impairments, and core-dependent amplifier variations. Conventional methods struggle with scalability and accuracy in multi-core, multi-span environments. To address these limitations, we propose the first digital twin framework for submarine SDM networks, integrating a neural-network-based model to capture nonlinear spatial–spectral interactions and physical topology constraints. Two launch power profile (LPP) optimization strategies are implemented—one for spectral flattening along single lightpaths, and another for maximizing and equalizing output power across multiple lightpaths. The digital twin (DT)-based optimization method, enhanced by an iterative refinement strategy and validated on a live-operated submarine optical network with 336 space-wavelength channels, outperformed traditional DT-only approaches by reducing error by 38% and improving stability by 39% in single-lightpath configurations, while achieving an 84% improvement in stability with comparable error in multi-lightpath scenarios. Integrated with a software-defined networking (SDN) control plane, the method supports real-time, automated LPP configuration without disrupting data traffic. The demonstrated accuracy, adaptability, and scalability of the proposed approach confirm its suitability for enabling efficient planning, control, and optimization in cognitive optical networks via intelligent power management for next-generation submarine SDM systems.