The decarbonization of the energy sector, driven by the growing use of renewable energy sources, is making power systems more variable, less predictable and harder to operate. One of the effects is grid congestion, particularly severe in the Netherlands, where it limits industrial electrification and the connection of new renewable capacity. A promising solution is to couple new loads with local generation and storage in micro-grids and multi-energy systems, modelled as energy hubs. Assessing their effectiveness requires accurate simulations, yet models always retain a residual error that can undermine their validity. Calibration corrects this error using measured data, but the large number of available methods, developed across different disciplines, makes their selection and use largely guided by experience. This thesis addresses this gap. After reviewing the sources of model error and the limits of calibration, calibration methods are classified into four categories: deterministic, stochastic, protocol-based and probabilistic. On this basis, a sequential decision framework is proposed, in which the discrepancy is progressively diagnosed and the calibration approach is selected according to the type of model, the nature of the error, the available data and the intended outcome. The framework is applied to four components of an energy hub: a wind turbine, a battery energy storage system, a residential photovoltaic plant and a wind forecast system, which differ in model, measured data and initial error. For all four, the calibrated models reached the defined accuracy requirements, proving the effectiveness of the proposed methodology. Given the breadth of the subject, the study had to be restricted to a limited set of methods and case studies, which places practical limits on the proposed methodology: it points to the suitable category, but the choice of the specific method and of its settings still relies on the experience of the user. Nevertheless, this work provides a solid basis for future studies towards a comprehensive methodology for calibration of energy systems model.
La decarbonizzazione del settore energetico, guidata dal crescente utilizzo di fonti rinnovabili, sta rendendo i sistemi energetici più variabili, meno prevedibili e più difficili da gestire. Uno degli effetti è la congestione di rete, particolarmente grave nei Paesi Bassi, dove limita l'elettrificazione dei processi industriali e la connessione di nuovi impianti rinnovabili. Una soluzione promettente consiste nell'accoppiare nuovi carichi con generazione e accumulo locali, in microreti e sistemi multi-energia modellati come energy hub. Valutarne l'efficacia richiede simulazioni accurate, ma i modelli mantengono sempre un errore residuo che può comprometterne la validità. La calibrazione corregge tale errore utilizzando dati misurati, ma l'elevato numero di metodi disponibili, sviluppati in discipline diverse, rende la loro scelta e uso largamente guidata dall'esperienza. In questo contesto si colloca il presente lavoro di tesi. Dopo un'analisi delle sorgenti di errore dei modelli e dei limiti della calibrazione, i metodi di calibrazione vengono classificati in quattro categorie: deterministici, stocastici, basati su protocolli e probabilistici. Sulla base di questa categorizzazione, viene proposta una struttura decisionale sequenziale che diagnostica la discrepanza e individua la calibrazione più adatta, a seconda del tipo di modello, della natura dell'errore, dei dati disponibili e del risultato desiderato. La struttura è applicata a quattro componenti di un energy hub: una turbina eolica, un sistema di accumulo a batterie, un impianto fotovoltaico residenziale e un sistema di previsione del vento, che differiscono per modello, dati misurati ed errore iniziale. In tutti e quattro i casi, i modelli calibrati hanno raggiunto i requisiti di accuratezza definiti, dimostrando l’efficacia della metodologia proposta. Data la vastità dell'argomento, lo studio è stato limitato a un insieme ristretto di metodi e casi studio, il che impone limiti pratici alla metodologia proposta: essa indica la categoria adatta, ma la scelta del metodo specifico e della sua configurazione resta affidata all’esperienza dell’utente. Ciononostante, questo lavoro costituisce una solida base per studi futuri di una metodologia completa per la calibrazione dei modelli di sistemi energetici.
Analyzing model-data mismatch: a decision framework for calibration of physics-based energy systems models
Di BERARDINO, NIKO
2025/2026
Abstract
The decarbonization of the energy sector, driven by the growing use of renewable energy sources, is making power systems more variable, less predictable and harder to operate. One of the effects is grid congestion, particularly severe in the Netherlands, where it limits industrial electrification and the connection of new renewable capacity. A promising solution is to couple new loads with local generation and storage in micro-grids and multi-energy systems, modelled as energy hubs. Assessing their effectiveness requires accurate simulations, yet models always retain a residual error that can undermine their validity. Calibration corrects this error using measured data, but the large number of available methods, developed across different disciplines, makes their selection and use largely guided by experience. This thesis addresses this gap. After reviewing the sources of model error and the limits of calibration, calibration methods are classified into four categories: deterministic, stochastic, protocol-based and probabilistic. On this basis, a sequential decision framework is proposed, in which the discrepancy is progressively diagnosed and the calibration approach is selected according to the type of model, the nature of the error, the available data and the intended outcome. The framework is applied to four components of an energy hub: a wind turbine, a battery energy storage system, a residential photovoltaic plant and a wind forecast system, which differ in model, measured data and initial error. For all four, the calibrated models reached the defined accuracy requirements, proving the effectiveness of the proposed methodology. Given the breadth of the subject, the study had to be restricted to a limited set of methods and case studies, which places practical limits on the proposed methodology: it points to the suitable category, but the choice of the specific method and of its settings still relies on the experience of the user. Nevertheless, this work provides a solid basis for future studies towards a comprehensive methodology for calibration of energy systems model.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_10_DiBerardino_Executive_Summary.pdf
accessibile in internet per tutti
Descrizione: Tesi di Laurea Magistrale
Dimensione
935.53 kB
Formato
Adobe PDF
|
935.53 kB | Adobe PDF | Visualizza/Apri |
|
2026_10_DiBerardino_Thesis.pdf
accessibile in internet per tutti
Descrizione: Executive Summary della Tesi di Laurea Magistrale: Analyzing Model-Data Mismatch: A Decision Framework for Calibration of Physics-Based Energy System Models
Dimensione
4.23 MB
Formato
Adobe PDF
|
4.23 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/266300