Modelling, Simulation and Use Cases for Digital Twin in Optical Networks

A comprehensive technical exploration of how digital twin technology is transforming optical network planning, operations, and optimization through physics-based models, machine learning, and real-time telemetry integration.

1.1 The Case for Digital Twins in Optical Networking

Optical transport networks form the backbone of global communications infrastructure, carrying over 95% of intercontinental data traffic across submarine cables and terrestrial fiber routes. As these networks evolve to support 400G, 800G, and beyond-terabit per-channel line rates, the complexity of managing the physical layer has grown dramatically. Operators must contend with a growing list of interacting impairments including amplified spontaneous emission (ASE) noise, chromatic dispersion (CD), polarization mode dispersion (PMD), polarization-dependent loss (PDL), fiber Kerr nonlinearities such as self-phase modulation (SPM), cross-phase modulation (XPM), and four-wave mixing (FWM), as well as stimulated Raman scattering (SRS) in wideband systems.

Traditional network planning tools rely on worst-case design margins, often allocating 3 to 6 dB of excess optical signal-to-noise ratio (OSNR) beyond the minimum required by the forward error correction (FEC) threshold. While these margins protect against unexpected degradation, they also waste significant capacity. A network that allocates 3 dB of unnecessary margin is effectively using only half of its potential throughput on each lightpath. With the cost of deploying new fiber routes measured in millions of dollars per kilometer for submarine cables and tens of thousands per kilometer for terrestrial routes, squeezing maximum capacity from existing infrastructure is a major economic driver.

A digital twin of an optical network addresses this challenge by creating a comprehensive virtual replica of the physical network. This replica incorporates models of every network element -- fiber span, amplifier, reconfigurable optical add-drop multiplexer (ROADM), and transponder -- and computes the expected signal quality for any configuration of wavelengths, power levels, and routing paths. Unlike static planning tools that compute worst-case scenarios, a network digital twin (NDT) continuously assimilates monitoring telemetry from the live network to refine its models and provide increasingly accurate predictions over time.

The concept of digital twins originated in manufacturing and aerospace, where virtual replicas of physical assets have been used for decades to predict wear, simulate performance, and optimize maintenance schedules. The application to optical networking presents unique challenges. Unlike mechanical systems where finite element analysis on static structures provides high accuracy, optical networks involve the propagation of electromagnetic fields through hundreds of kilometers of fiber, with nonlinear interactions between tens or hundreds of wavelength channels simultaneously. The accuracy of the digital twin depends on the fidelity of models for each individual component and on how those models compose when cascaded across a network path.

1.2 Scope and Organization

This article provides a comprehensive technical treatment of digital twin technology for optical networks. It covers the historical evolution from static planning tools to dynamic digital twins, the theoretical framework based on the Gaussian Noise (GN) model and its extensions, the architecture and component models that comprise a practical optical NDT, machine learning techniques that enhance model accuracy, open-source implementations such as GNPy, and the integration of digital twins with software-defined networking (SDN) controllers and real-time telemetry systems. Practical use cases demonstrated in field trials and research testbeds are analyzed, followed by a quantitative discussion of accuracy benchmarks and computational performance. The article concludes with an assessment of current challenges and a forward-looking perspective on how digital twin technology will evolve alongside multiband transmission, autonomous network operation, and AI-driven optimization.

2.1 Evolution of Network Planning Tools

The history of optical network design tools mirrors the evolution of optical transmission technology itself. In the early days of wavelength-division multiplexing (WDM) systems in the mid-1990s, network planning was performed using spreadsheet-based link budget calculations. Engineers would sum the losses from fiber attenuation, connector losses, splice losses, and ROADM insertion losses, then verify that the received OSNR met the requirements for the chosen modulation format -- typically non-return-to-zero on-off keying (NRZ-OOK) at 2.5 or 10 Gb/s. These calculations assumed worst-case parameters for every component and treated each span independently.