AEIT International Conference, Rome, 2022 © 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. The Role of Digital Twins in Power System Inertia Estimation Fabrizio De Caro Department of Engineering University of Sannio Benevento, Italy fdecaro@unisannio.it Viktoriya Mostova Department of Engineering University of Sannio Benevento, Italy v.mostova@studenti.unisannio.it Alfredo Vaccaro Department of Engineering University of Sannio Benevento, Italy vaccaro@unisannio.it Abstract—Modern power systems are experiencing a deep transformation phase, as a result of the increasing penetration of renewable power generators, which causes many consequences on grid stability. In this scenario, power system inertia is rapidly decreasing and extremely variable, pushing system operators to develop reliable tools enabling online inertia estimation. To effectively address this challenge, system operators could develop a mirrored copy of the system, called Digital Twin, which allows performing advanced online analyses aimed at studying the dynamic behavior of the grid. To outline the potential role of this emerging computing paradigm in the context of power system dynamics, this paper analyzes the performance of adaptive data- driven models in online grid parameter estimation. A two-area model is considered, where the experimental results showed the effectiveness of the analyzed methods in reliably reproducing the frequency evolution under different operation scenarios. Index Terms—Digital Twin, Power System Operation, Inertia Estimation, ARMAX, Data-Driven Models. I. I NTRODUCTION The need for power systems decarbonization is inducing a widespread penetration of renewable energy sources (RES) in existing electrical grids. In particular, [1] reports an average annual (worldwide) growth rate of 15% for photovoltaics and 10% for wind generators, hence increasing the RES contribution to the overall generation mix [2]. In this complex scenario, one of the most challenging problem to face is the continuing reduction of the mechanical inertia provided by Synchronous Generators (SG), which is defined as the ratio between the kinetic energy and the generator rated power. System inertia represents a valuable resource aimed at counteracting frequency variations in the very first moments after a power imbalance, long before pri- mary frequency control (PFC) is activated [1]. This important feature does not characterize renewable generators, which are frequently connected to the grid by power converters- based interfaces that decouple the generator dynamics from the grid frequency. Consequently, the replacement of SG by RES poses major challenges to frequency regulation, such as higher and faster Rates of Change of Frequency (RoCoF), which could result in unintentional tripping of over/under- frequency relays [3]. Such fast variations in RoCoF are dan- gerous because they reduce the time that generator controllers have to intervene before the frequency reaches trip thresholds causing untimely tripping of protection systems and reducing the effectiveness of load shedding techniques [1]. Several approaches have been proposed in the literature for trying and solving this complex issue. The main idea is to allow inverter-based generators to support the grid by providing frequency regulation services, which include fast frequency regulation and virtual inertia [4]. The effective deployment of these methods requires an on- line estimation of the actual power system inertia, so as to support system operators in identifying reliable control actions aimed at improving system stability and security, mitigating the effects of the large uncertainties induced by RES [5]. To solve this problem the deployment of Digital Twins (DTs)-based computing paradigms could play an important role. DT can be defined as “a set of virtual information constructs that mimics the structure, context and behavior of an individual or unique physical asset, that is dynamically up- dated with data from its physical twin throughout its life-cycle, and that ultimately informs decisions that realize value” [6]. Another interesting definition defines DT as “an integrated multi-physics, multi-scale, probabilistic simulation of a com- plex product and uses the best available physical models, sensor updates, etc., to mirror the life of its corresponding twin” [7]. Hence, DT is able to emulate the dynamic of complex systems in real operation scenario by properly amalgamating the information coming from both physical and data-driven models [8]. This feature is particularly useful in power system inertia estimation, where the availability of massive data- streams generated by time-synchronized grid sensors leads to