Geological storage of gases in deep saline aquifers is a key technology for climate change mitigation, both for permanent CO₂ sequestration and for large-scale underground hydrogen storage (UHS) in future low-carbon energy systems. Reliable predictions of gas solubility in brine under high-pressure conditions are essential for assessing storage capacity and long-term containment security. This work evaluates and compares three thermodynamic modelling approaches — Søreide–Whitson (modified Peng–Robinson), Electrolyte NRTL (eNRTL) and Pitzer — for predicting CO₂ solubility in pure water and NaCl brines under geological conditions (283–423 K, 10–400 bar, up to 6 molal NaCl). The Søreide–Whitson model is also evaluated for H₂ solubility in both pure water and brines under the same conditions. A rigorous data selection methodology was applied to a comprehensive experimental database to ensure robust model calibration and validation. For CO₂ systems, both eNRTL and Pitzer demonstrate excellent accuracy, with mean absolute average deviations (AAD) below 3.5 × 10⁻⁴ (salt free mole fraction) and average mean absolute percentage error (MAPE) values of 2.43% and 2.75%, respectively, for the ternary system. These metrics were calculated using the selected experimental datasets, excluding those previously identified as systematically inconsistent. The Søreide–Whitson model, in contrast, shows a slight tendency to underestimate solubility at low pressures and overestimate it at high pressures. Regarding computational performance, Pitzer emerges as computationally more efficient than eNRTL, requiring only 53% of the time. Furthermore, the models were compared against an Aspen Plus® V14 default version of the eNRTL model (i.e., with the default binary interaction parameters) to assess the impact of parameter fitting relative to the available tools. For H₂ systems, standard eNRTL and Pitzer implementations fail to capture solubility trends; only the Søreide–Whitson model yields acceptable results for the binary system, while performance deteriorates significantly in brine and remains strongly dataset-dependent. The practical implications are quantified through a case study on the Sleipner project (North Sea), where differences in model accuracy translate into variations of up to 60 kt in predicted CO₂ dissolution over 200 years. These findings emphasize the critical importance of rigorous model selection, high-quality experimental datasets, explicit uncertainty quantification, and the need for additional high-pressure H₂–brine measurements.
LLo stoccaggio geologico di gas in acquiferi salini profondi è una tecnologia chiave per la mitigazione del cambiamento climatico, sia per il sequestro permanente di CO₂ che per lo stoccaggio sotterraneo di idrogeno su larga scala (UHS) nei futuri sistemi energetici a basse emissioni di carbonio. Previsioni affidabili della solubilità dei gas in salamoia in condizioni di alta pressione sono essenziali per valutare la capacità di stoccaggio e la sicurezza del confinamento a lungo termine. Questo lavoro valuta e confronta tre approcci di modellazione termodinamica — Søreide–Whitson (Peng–Robinson modificato), Electrolyte NRTL (eNRTL) e Pitzer — per la previsione della solubilità della CO₂ in acqua pura e in salamoie NaCl in condizioni geologiche (283–423 K, 10–400 bar, fino a 6 molali NaCl). Il modello Søreide–Whitson è valutato anche per la solubilità dell'H₂ nelle stesse condizioni. Una rigorosa metodologia di selezione dei dati è stata applicata a un ampio database sperimentale per garantire una calibrazione e validazione robusta. Per i sistemi CO₂, eNRTL e Pitzer mostrano accuratezza eccellente, con deviazioni assolute medie (AAD) inferiori a 3.5 × 10⁻⁴ (frazione molare sal free) e valori medi dell'errore percentuale assoluto (MAPE) del 2.43% e 2.75% per il sistema ternario, escludendo i dataset precedentemente identificati come sistematicamente incoerenti. Søreide–Whitson tende invece a sottostimare a basse pressioni e sovrastimare a pressioni elevate. Dal punto di vista computazionale, Pitzer risulta più efficiente di eNRTL, richiedendo solo il 53% del tempo. I modelli sono stati inoltre confrontati con la versione predefinita di eNRTL in Aspen Plus® V14 per valutare l'impatto della calibrazione dei parametri. Per l'H₂, le implementazioni standard di eNRTL e Pitzer non catturano gli andamenti di solubilità; solo Søreide–Whitson fornisce risultati accettabili per il sistema binario, mentre in salamoia le prestazioni peggiorano e rimangono fortemente dipendenti dal dataset. Le implicazioni pratiche sono quantificate attraverso un caso di studio sul progetto Sleipner (Mare del Nord), dove le differenze nei modelli si traducono in variazioni fino a 60 kt nella CO₂ disciolta su 200 anni. Questi risultati sottolineano l'importanza critica di una rigorosa selezione dei modelli e dei dati sperimentali, della quantificazione esplicita dell'incertezza e della necessità di ulteriori misure di H₂ in salamoia ad alta pressione.
Evaluation of thermodynamic models for CO2 and H2 solubility in pure water and brines under geological conditions
CALLERI GELOS, DIEGO
2025/2026
Abstract
Geological storage of gases in deep saline aquifers is a key technology for climate change mitigation, both for permanent CO₂ sequestration and for large-scale underground hydrogen storage (UHS) in future low-carbon energy systems. Reliable predictions of gas solubility in brine under high-pressure conditions are essential for assessing storage capacity and long-term containment security. This work evaluates and compares three thermodynamic modelling approaches — Søreide–Whitson (modified Peng–Robinson), Electrolyte NRTL (eNRTL) and Pitzer — for predicting CO₂ solubility in pure water and NaCl brines under geological conditions (283–423 K, 10–400 bar, up to 6 molal NaCl). The Søreide–Whitson model is also evaluated for H₂ solubility in both pure water and brines under the same conditions. A rigorous data selection methodology was applied to a comprehensive experimental database to ensure robust model calibration and validation. For CO₂ systems, both eNRTL and Pitzer demonstrate excellent accuracy, with mean absolute average deviations (AAD) below 3.5 × 10⁻⁴ (salt free mole fraction) and average mean absolute percentage error (MAPE) values of 2.43% and 2.75%, respectively, for the ternary system. These metrics were calculated using the selected experimental datasets, excluding those previously identified as systematically inconsistent. The Søreide–Whitson model, in contrast, shows a slight tendency to underestimate solubility at low pressures and overestimate it at high pressures. Regarding computational performance, Pitzer emerges as computationally more efficient than eNRTL, requiring only 53% of the time. Furthermore, the models were compared against an Aspen Plus® V14 default version of the eNRTL model (i.e., with the default binary interaction parameters) to assess the impact of parameter fitting relative to the available tools. For H₂ systems, standard eNRTL and Pitzer implementations fail to capture solubility trends; only the Søreide–Whitson model yields acceptable results for the binary system, while performance deteriorates significantly in brine and remains strongly dataset-dependent. The practical implications are quantified through a case study on the Sleipner project (North Sea), where differences in model accuracy translate into variations of up to 60 kt in predicted CO₂ dissolution over 200 years. These findings emphasize the critical importance of rigorous model selection, high-quality experimental datasets, explicit uncertainty quantification, and the need for additional high-pressure H₂–brine measurements.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_03_Calleri_Diego_Executive_Summary.pdf
non accessibile
Dimensione
1.51 MB
Formato
Adobe PDF
|
1.51 MB | Adobe PDF | Visualizza/Apri |
|
2026_03_Calleri_Diego_Manuscript.pdf
non accessibile
Dimensione
4.91 MB
Formato
Adobe PDF
|
4.91 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/10589/253667