The simulation of complex subsurface fluid flow dynamics, such as those involved in Carbon Capture an Sequestration (CCS), requires accurate mathematical models and robust numerical methods. CCS, a promising technique for mitigating climate change, involves capturing carbon dioxide from ambient air or industrial processes and injecting it into subsurface geological formations. Traditional simulation approaches often result in a compromise between model complexity and numerical accuracy. Our interest is to address this challenge by employing a high-accuracy method to simulate a comprehensive model of the CCS process. An appropriate mathematical representation of the CCS process, together with an accurate simulation of the system, is fundamental in order to understand the safety of the storage and the long-term impact of these techniques. Given the importance of the mathematical model, the work done in this thesis begins with the derivation of a system of equations that describes the CCS process with mild simplifying assumptions. This leads us to derive a coupled system of highly non-linear partial differential equations (PDEs) with a mixed hyperbolic and parabolic behavior. In the existing literature to simulate this kind of complex systems, low-order methods are generally employed, such as integral finite difference and finite volume methods. Our objective is to use a high-accuracy method to simulate the full system. In order to achieve this, we employ a splitting technique, which had previously been applied to a simplified version of the model. The decoupling that we choose for the full CCS system is achieved through the addition and subtraction of an elliptic term that depends only on the pressure and not on the saturation. The decoupled system is discretized in space using a combination of upwind and Symmetric Interior Penalty (SIP) Discontinuous Galerkin methods. The upwind scheme is employed for the hyperbolic term, while SIP is used for the parabolic term. The time discretization is handled using the theta-method. The resulting non-linear system is then solved through Newton's method, which does not converge in this case. One of the potential causes for which the scheme does not converge can be attributed to the presence of boundary integrals of non-linear terms. These terms are numerically challenging to treat and they arise from the use of SIP. In order to address this issue, as an alternative DG approach for the parabolic term, we consider the Local Discontinuous Galerkin method (LDG). This scheme, through the rewriting of the system in a mixed formulation, eliminates the boundary terms. The thesis aims to determine whether LDG method can be employed to solve the parabolic term of the decoupled system of the full CCS model. To this end, we consider non-linear cross-diffusion models, which exhibit a similar structure to the parabolic term of the full model. In order to assess this, we consider a variety of models as benchmarks, including the Shigesada-Kawasaki-Teramoto population model, the Busenberg-Travis population model and the three-component Maxwell-Stefan system for gas mixtures. The results of our numerical experiments show that the LDG scheme typically exhibits better performance when compared to the numerical methods that are employed in the existing literature, often improving computational costs or convergence rates. This provides valuable insight into the potential of using the combination of upwind and LDG, to simulate the complete system. Our future work will be focused on identifying the optimal splitting of the system, which remains an open question. Then, we intend to apply the LDG scheme instead of the SIP for the treatment of the parabolic term, with the aim of bridging the gap between model completeness and the accuracy of the method.
La simulazione delle complesse dinamiche di flusso di fluidi nel sottosuolo, come quelle coinvolte nei processi di cattura e stoccaggio del carbonio (CCS), richiede modelli matematici accurati e metodi numerici robusti. La CCS, una tecnica promettente per mitigare i cambiamenti climatici, consiste nel catturare l’anidride carbonica dall'aria o dai processi industriali e iniettarla in formazioni geologiche sotterranee. Gli approcci di simulazione tradizionali spesso comportano un compromesso tra la complessità del modello e l'accuratezza numerica. Il nostro obiettivo è affrontare questa sfida utilizzando un metodo ad alta accuratezza per simulare un modello completo della CCS. Una rappresentazione matematica adeguata della CCS, insieme a una simulazione accurata del sistema, è fondamentale per comprendere la sicurezza dello stoccaggio e l'impatto a lungo termine di queste tecniche. Data l’importanza del modello matematico, il lavoro svolto in questa tesi inizia con la derivazione di un sistema di equazioni che descrive la CCS con ipotesi semplificative. Questo ci porta a derivare un sistema accoppiato di PDE altamente non lineari con un comportamento misto iperbolico e parabolico. Nella letteratura esistente, per simulare questo tipo di sistemi complessi, vengono generalmente utilizzati metodi di ordine basso, come il metodo delle differenze finite integrali e il metodo dei volumi finiti. Il nostro obiettivo è utilizzare un metodo ad alta accuratezza per simulare l'intero sistema. Per raggiungere questo obiettivo, applichiamo una tecnica di splitting che era stata precedentemente utilizzata per una versione semplificata del modello. Lo splitting che scegliamo per il sistema completo di CCS si ottiene attraverso l'aggiunta e la sottrazione di un termine ellittico che dipende solo dalla pressione e non dalla saturazione. Il sistema separato viene discretizzato in spazio utilizzando una combinazione di metodi Discontinuous Galerkin con schema upwind per il termine iperbolico e Symmetric Interior Penalty (SIP) per il termine parabolico. La discretizzazione temporale viene gestita con il theta-metodo. Il sistema non lineare risultante viene poi risolto tramite il metodo di Newton, che in questo caso non converge. Una delle cause potenziali per cui lo schema non converge è attribuibile alla presenza di integrali di bordo di termini non lineari, numericamente difficili da trattare, che emergono dall'uso del metodo SIP. Per affrontare questo problema, consideriamo un approccio DG alternativo per il termine parabolico: il metodo Local Discontinuous Galerkin (LDG). Questo schema, riscrivendo il sistema in formulazione mista, elimina i termini di bordo. La tesi mira a determinare se il metodo LDG può essere impiegato per risolvere il termine parabolico del sistema splittato nel modello completo di CCS. A tal fine, consideriamo modelli di cross-diffusion non lineari, che presentano una struttura simile al termine parabolico del modello completo. Per valutare questo, utilizziamo una serie di modelli come benchmark, tra cui il modello di popolazione di Shigesada-Kawasaki-Teramoto, il modello di popolazione di Busenberg-Travis e il sistema a tre componenti di Maxwell-Stefan per miscele gassose. I risultati dei nostri esperimenti numerici mostrano che lo schema LDG tende a offrire prestazioni migliori rispetto ai metodi numerici utilizzati nella letteratura esistente, migliorando spesso i costi computazionali o i rate di convergenza. Questo fornisce preziose indicazioni sul potenziale utilizzo della combinazione di upwind e LDG per simulare il sistema completo. Il nostro lavoro futuro si concentrerà sull'identificazione dello splitting ottimale del sistema, che rimane una questione aperta. Successivamente, intendiamo applicare lo schema LDG al posto del SIP per il trattamento del termine parabolico, con l'obiettivo di colmare il divario tra la completezza del modello e l'accuratezza del metodo.
Exploring Local Discontinuous Galerkin methods for cross-diffusion systems: Towards complex multi-phase flow simulation
LOCATELLI, FRANCESCA
2025
Abstract
The simulation of complex subsurface fluid flow dynamics, such as those involved in Carbon Capture an Sequestration (CCS), requires accurate mathematical models and robust numerical methods. CCS, a promising technique for mitigating climate change, involves capturing carbon dioxide from ambient air or industrial processes and injecting it into subsurface geological formations. Traditional simulation approaches often result in a compromise between model complexity and numerical accuracy. Our interest is to address this challenge by employing a high-accuracy method to simulate a comprehensive model of the CCS process. An appropriate mathematical representation of the CCS process, together with an accurate simulation of the system, is fundamental in order to understand the safety of the storage and the long-term impact of these techniques. Given the importance of the mathematical model, the work done in this thesis begins with the derivation of a system of equations that describes the CCS process with mild simplifying assumptions. This leads us to derive a coupled system of highly non-linear partial differential equations (PDEs) with a mixed hyperbolic and parabolic behavior. In the existing literature to simulate this kind of complex systems, low-order methods are generally employed, such as integral finite difference and finite volume methods. Our objective is to use a high-accuracy method to simulate the full system. In order to achieve this, we employ a splitting technique, which had previously been applied to a simplified version of the model. The decoupling that we choose for the full CCS system is achieved through the addition and subtraction of an elliptic term that depends only on the pressure and not on the saturation. The decoupled system is discretized in space using a combination of upwind and Symmetric Interior Penalty (SIP) Discontinuous Galerkin methods. The upwind scheme is employed for the hyperbolic term, while SIP is used for the parabolic term. The time discretization is handled using the theta-method. The resulting non-linear system is then solved through Newton's method, which does not converge in this case. One of the potential causes for which the scheme does not converge can be attributed to the presence of boundary integrals of non-linear terms. These terms are numerically challenging to treat and they arise from the use of SIP. In order to address this issue, as an alternative DG approach for the parabolic term, we consider the Local Discontinuous Galerkin method (LDG). This scheme, through the rewriting of the system in a mixed formulation, eliminates the boundary terms. The thesis aims to determine whether LDG method can be employed to solve the parabolic term of the decoupled system of the full CCS model. To this end, we consider non-linear cross-diffusion models, which exhibit a similar structure to the parabolic term of the full model. In order to assess this, we consider a variety of models as benchmarks, including the Shigesada-Kawasaki-Teramoto population model, the Busenberg-Travis population model and the three-component Maxwell-Stefan system for gas mixtures. The results of our numerical experiments show that the LDG scheme typically exhibits better performance when compared to the numerical methods that are employed in the existing literature, often improving computational costs or convergence rates. This provides valuable insight into the potential of using the combination of upwind and LDG, to simulate the complete system. Our future work will be focused on identifying the optimal splitting of the system, which remains an open question. Then, we intend to apply the LDG scheme instead of the SIP for the treatment of the parabolic term, with the aim of bridging the gap between model completeness and the accuracy of the method.File | Dimensione | Formato | |
---|---|---|---|
phd_unimib_874114.pdf
accesso aperto
Dimensione
2.78 MB
Formato
Adobe PDF
|
2.78 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/20.500.14242/201102
URN:NBN:IT:UNIMIB-201102