The rapid growth of inverter-based resources (IBRs), particularly wind, solar and grid-scale battery storage is introducing new stability challenges such as poorly damped oscillations that are difficult to foresee and mitigate using conventional tools. This paper presents a frequency-domain digital twin (DT) of an IBR-dominated power system for real-time situational awareness and early warning of such oscillations in a control room.

The proposed DT focuses on identifying emerging poorly damped oscillations, their root cause and geographical spread, enabling preventive mitigation action near real time. A key challenge in developing such a DT is the lack of transparency of vendor-specific IBR models. Estimated IBR transfer functions (TFs) from dynamic frequency scans (DFS) are tied to specific operating points, making them inadequate for systems undergoing frequent redispatch. Repeating DFS on a wide-area EMT model every time the operating point changes is prohibitively slow and demands substantial resources, making it unsuitable for real-time applications. Furthermore, ensuring appropriate perturbation amplitudes for system-wide DFS without triggering nonlinearities such as current limits or transformer saturation adds another layer of complexity.

To overcome these, the DT reported here adopts a modular bottom-up approach that is tractable and scalable. Instead of relying on DFS in real time, a library of estimated TFs for each IBR plant is constructed offline covering its entire range of operating points (e.g., loading levels). These can be mandated via grid code before commissioning. In real time, the TF of each IBR plant corresponding to the prevailing operating point is interpolated from those in the library using a novel geometric clustering technique and regression. This enables accurate representation of IBR dynamics without repeated DFS.

By combining the TFs of individual IBR plants at the prevailing operating point together with known passive network and other component models (synchronous machines, loads etc.), the overall system TF (or an equivalent state-space) is dynamically updated as a frequency-domain DT. This enables continuous tracking of system dynamics as operating conditions evolve and supports modal analysis for early detection of poorly damped oscillatory modes. Crucially, it also facilitates pinpointing dominant contributors using participation factors and reveals the spatial distribution of oscillations through modal observability, thereby enhancing operator insight and preventive mitigation.

Beyond real-time operation, the proposed DT has broader applications in planning and compliance. It can significantly reduce the need for computationally intensive DFS on wide-area EMT models and excessively high-order vector fitting for system-wide stability studies, offering a faster and more reliable alternative for IBR connection studies and scenario analysis.

The full paper will present: (i) the architecture and step-by-step development of the DT, (ii) its ability to track operating point variations and provide early warnings under system redispatch, (iii) its capability to identify key contributors and map oscillation propagation across the grid. The findings will be demonstrated on a modified IEEE test system with very high IBR penetration along with illustrative control room visualisations and (iv) analysis of computation effort and time to establish implementation feasibility in real time.