The increasing penetration of Inverter-Based Resources (IBRs) in modern power systems is progressively displacing conventional synchronous generation, reducing system inertia and posing new challenges for frequency stability and dynamic performance. Grid-Forming (GFM) converters have emerged as a solution to these challenges, as they are capable of replicating grid support functionalities traditionally associated with synchronous machines. This thesis investigates the control, modeling, operation, and system identification of GFM converters, with particular focus on frequency support provision, DC-side source interaction, and terminal parameter estimation from measured data. As a theoretical preamble to frequency and inertia support, a control strategy for upward power reserve provision by grid-connected Photovoltaic (PV) plants without dedicated energy storage is developed. The plant operates at a Flexible Power Point (FPP), determined through a piecewise linear approximation of the power--voltage characteristic, enabling frequency containment reserve comparable to conventional droop control. This work provides an entry point to the broader investigation of frequency and inertia support in GFM converters developed throughout the thesis. The thesis then develops full-order small-signal state-space models of GFM converters from first principles via the Component Connection Method (CCM), constituting a white-box modeling approach. These models form the basis for the design and analysis of a Virtual Synchronous Generator (VSG)-based control strategy and are experimentally validated through a Power Hardware-in-the-Loop (PHIL) platform. The influence of DC-side configuration on frequency support provision through virtual inertia is systematically examined across three cases: an ideal DC source, a Battery Energy Storage System (BESS), and a wind turbine-based source. Each is evaluated through the Rate of Change of Frequency (RoCoF) and frequency nadir metrics and benchmarked against a synchronous generator of equivalent inertia. The analysis is further extended to Permanent Magnet Synchronous Generator (PMSG) wind turbines operating under GFM-VSG control, comparing Maximum Power Point Tracking (MPPT) and Deloaded operating modes, and assessing the effect of frequency droop on inertial response. Finally, a grey-box system identification framework is developed and experimentally validated on a real grid-connected GFM converter. The framework is implemented using nonlinear least-squares optimization applied to Reduced-Order Models (ROMs) derived from the full-order white-box model, considering three reduction techniques: Balanced Truncation (BT), Singular Perturbation with Participation Factors (SP-PF), and Singular Perturbation with Energy Residuals (SP-Energy). This is complemented by a structured identification approach using MATLAB's idgrey environment, in which different numerical estimators, including Gauss--Newton and Levenberg--Marquardt methods, are benchmarked for convergence and robustness. A curve-fitting-based implementation is additionally considered as a practical reference. In parallel, a black-box modeling approach based on the Eigensystem Realization Algorithm (ERA) is used to derive ROMs directly from input--output measurements. The results highlight the trade-off between time-domain accuracy and preservation of modal properties, with direct implications for small-signal stability analysis. Together, these contributions provide both theoretical foundations and practical tools for the deployment of GFM converters in future low-inertia power systems.
La crescente penetrazione delle Risorse Basate su Inverter (Inverter-Based Resources, IBR) nei moderni sistemi elettrici sta progressivamente sostituendo la generazione sincrona convenzionale, riducendo l'inerzia di sistema e introducendo nuove sfide per la stabilit\`a di frequenza e sulle prestazioni dinamiche. I convertitori grid-forming (GFM) sono emersi come soluzione a queste sfide, poich\'e sono in grado di replicare le funzionalit\`a di supporto alla rete tradizionalmente associate alle macchine sincrone. Questa tesi analizza il controllo, la modellazione e il funzionamento dei convertitori GFM, con particolare attenzione al supporto della frequenza, all'interazione con la sorgente sul lato DC e all'identificazione del sistema. Come premessa teorica al supporto di frequenza e inerzia, viene sviluppata una strategia di controllo per la fornitura di riserva di potenza verso l'alto da parte di impianti fotovoltaici (PV) connessi alla rete. L'impianto opera in un Punto di Potenza Flessibile (Flexible Power Point, FPP), determinato tramite un'approssimazione lineare a tratti della caratteristica potenza--tensione, che consente una riserva di contenimento della frequenza paragonabile al controllo droop convenzionale. Questo lavoro costituisce un punto di accesso all'indagine pi\`u ampia sul supporto di frequenza e inerzia nei convertitori GFM sviluppata nel resto della tesi. La tesi sviluppa quindi modelli a stato di piccolo segnale a ordine completo dei convertitori GFM a partire dai principi primi mediante il Component Connection Method (CCM), costituendo un approccio di modellazione white-box. Questi modelli costituiscono la base per la progettazione e l'analisi di una strategia di controllo basata su Generatore Sincrono Virtuale (Virtual Synchronous Generator, VSG) e sono validati sperimentalmente tramite una piattaforma Power Hardware-in-the-Loop (PHIL). L'influenza della configurazione del lato DC sulla fornitura di supporto di frequenza tramite inerzia virtuale \`e esaminata sistematicamente su tre casi: una sorgente DC ideale, un sistema di accumulo a batteria (BESS) e una sorgente basata su turbina eolica. Ciascun caso \`e valutato tramite le metriche di Rate of Change of Frequency (RoCoF) e di nadir di frequenza, confrontandolo con un generatore sincrono di inerzia equivalente. L'analisi \`e successivamente estesa a turbine eoliche con Generatore Sincrono a Magneti Permanenti (PMSG) operanti sotto controllo GFM-VSG, confrontando le modalit\`a Maximum Power Point Tracking (MPPT) e Deloaded, e valutando l'effetto del droop di frequenza sulla risposta inerziale. Infine, viene sviluppato e validato sperimentalmente su un convertitore GFM reale connesso in rete un framework di identificazione grey-box. Il framework \`e implementato mediante ottimizzazione ai minimi quadrati non lineare applicata a Modelli a Ordine Ridotto (ROM) derivati dal modello white-box a ordine completo, considerando tre tecniche di riduzione: Balanced Truncation (BT), Singular Perturbation with Participation Factors (SP-PF) e Singular Perturbation with Energy Residuals (SP-Energy). Questo approccio \`e integrato da una procedura di identificazione strutturata che utilizza l'ambiente \texttt{idgrey} di MATLAB, in cui diversi stimatori numerici, tra cui i metodi di Gauss--Newton e Levenberg--Marquardt, vengono confrontati in termini di convergenza e robustezza. Viene inoltre considerata, come riferimento pratico, un'implementazione basata su curve-fitting. In parallelo, un approccio di modellazione black-box basato sull'Eigensystem Realization Algorithm (ERA) \`e utilizzato per derivare i ROM direttamente da misure di ingresso--uscita. I risultati evidenziano il compromesso tra precisione nel dominio del tempo e preservazione delle propriet\`a modali, con implicazioni dirette per l'analisi di stabilit\`a al piccolo segnale. Nel loro insieme, questi contributi forniscono sia fondamenti teorici sia strumenti pratici per la diffusione dei convertitori GFM nei futuri sistemi elettrici a bassa inerzia.
Grid-Forming Converters: Control, Modeling, Operation, and System Identification in Inverter-Based Power Systems
CARDENAS CALLE, MARCELO GUSTAVO
2026
Abstract
The increasing penetration of Inverter-Based Resources (IBRs) in modern power systems is progressively displacing conventional synchronous generation, reducing system inertia and posing new challenges for frequency stability and dynamic performance. Grid-Forming (GFM) converters have emerged as a solution to these challenges, as they are capable of replicating grid support functionalities traditionally associated with synchronous machines. This thesis investigates the control, modeling, operation, and system identification of GFM converters, with particular focus on frequency support provision, DC-side source interaction, and terminal parameter estimation from measured data. As a theoretical preamble to frequency and inertia support, a control strategy for upward power reserve provision by grid-connected Photovoltaic (PV) plants without dedicated energy storage is developed. The plant operates at a Flexible Power Point (FPP), determined through a piecewise linear approximation of the power--voltage characteristic, enabling frequency containment reserve comparable to conventional droop control. This work provides an entry point to the broader investigation of frequency and inertia support in GFM converters developed throughout the thesis. The thesis then develops full-order small-signal state-space models of GFM converters from first principles via the Component Connection Method (CCM), constituting a white-box modeling approach. These models form the basis for the design and analysis of a Virtual Synchronous Generator (VSG)-based control strategy and are experimentally validated through a Power Hardware-in-the-Loop (PHIL) platform. The influence of DC-side configuration on frequency support provision through virtual inertia is systematically examined across three cases: an ideal DC source, a Battery Energy Storage System (BESS), and a wind turbine-based source. Each is evaluated through the Rate of Change of Frequency (RoCoF) and frequency nadir metrics and benchmarked against a synchronous generator of equivalent inertia. The analysis is further extended to Permanent Magnet Synchronous Generator (PMSG) wind turbines operating under GFM-VSG control, comparing Maximum Power Point Tracking (MPPT) and Deloaded operating modes, and assessing the effect of frequency droop on inertial response. Finally, a grey-box system identification framework is developed and experimentally validated on a real grid-connected GFM converter. The framework is implemented using nonlinear least-squares optimization applied to Reduced-Order Models (ROMs) derived from the full-order white-box model, considering three reduction techniques: Balanced Truncation (BT), Singular Perturbation with Participation Factors (SP-PF), and Singular Perturbation with Energy Residuals (SP-Energy). This is complemented by a structured identification approach using MATLAB's idgrey environment, in which different numerical estimators, including Gauss--Newton and Levenberg--Marquardt methods, are benchmarked for convergence and robustness. A curve-fitting-based implementation is additionally considered as a practical reference. In parallel, a black-box modeling approach based on the Eigensystem Realization Algorithm (ERA) is used to derive ROMs directly from input--output measurements. The results highlight the trade-off between time-domain accuracy and preservation of modal properties, with direct implications for small-signal stability analysis. Together, these contributions provide both theoretical foundations and practical tools for the deployment of GFM converters in future low-inertia power systems.| File | Dimensione | Formato | |
|---|---|---|---|
|
phdunige_5763946.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
21.97 MB
Formato
Adobe PDF
|
21.97 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/385514
URN:NBN:IT:UNIGE-385514