This thesis focuses on the optimal design and operation of Aggregate Energy Systems under uncertainty. These systems are composed by different dispatchable storage and generation units, as well as non-dispatchable, distributed energy sources. They are designed with the purpose of integrating different energy technologies so to improve the overall efficiency, meet different energy demands and facilitate the penetration of renewable energy sources. Much interest is nowadays present for these systems, since they are already economically competitive in certain applications (e.g., serving heat, power and cooling for industrial and urban users) and expected to play a major role in the energy sector. In literature different kinds of Aggregate Energy Systems can be found: Multi Energy Systems, microgrids, Virtual Power Plants, etc. Virtual Power Plant differentiate themselves from the rest since they are designed with the main goal to participate in the electricity markets, and they usually feature a larger installed power. Given the presence of different technologies and the complexity of the system, both the optimal design and operation require the development of mathematical tools able to capture the different technical aspects of each unit as well as the uncertain nature of the parameters (e.g., PV generation). In this Thesis, four topics of industrial interest are considered and analyzed, and four different approaches are proposed. These are: (1) a Rolling Horizon approach for the optimal robust operation of Aggregate Energy Systems featuring yearly constraints and/or seasonal storage; (2) a multi-stage stochastic optimization model for the optimal operation of Virtual Power Plants participating in the electricity markets; (3) a two-stage stochastic optimization model for the optimal design of Aggregate Energy Systems with (N-1)-reliability; (4) a methodology for the optimal design of zero carbon emission Virtual Power Plants with hydrogen seasonal storage. When applied to real-world case studies, the approaches resulted successful in the tasks they were designed for. At first, the optimal robust operation evaluated by the Rolling Horizon (1) was able to serve the user’s energy demands through an entire year and for all the case studies considered. In addition, the considered yearly constraints were met and the daily management of the seasonal storage proved to be successful in avoiding shortages. Regarding the study on the Virtual Power Plant (2), the multi-stage stochastic optimization model was effective in carefully modelling both the energy systems and the electricity markets, and the solutions obtained proved to be better than the ones obtained with a deterministic approach. Moreover, the computational time was significantly improved by developing a shrinking horizon-based decomposition, making the methodology suitable for practical industrial use. For what concerns the third research topic addressed in this thesis (3), the two-stage stochastic model for the optimal design of Aggregate Energy Systems with (N-1)-reliability was able to compute design solutions able to meet the users’ demands along a whole yearly horizon, for whichever contingency. The high solution time required the development of decomposition techniques, one of them very effective in shortening the computational time although not guaranteeing optimality. Finally, the optimal design of a future Virtual Power Plant featuring hydrogen storage and hydrogen-fired combine cycles was evaluated (4). The proposed methodology combines some of the features presented in the previous works, such as many parts of two-stage stochastic model of (3), the reserve constraints present in (2) and a modified version of the Rolling Horizon for the optimal schedule of the plant and seasonal storage management presented in (1). Results proved these concepts and evaluateed the techno-economic feasibility of such aggregate system. In conclusion, the methodologies presented in this Thesis were proven to be effective for the considered real-world case studies, although improvements can be made in terms of computational time and ease of implementation.
Questa tesi si focalizza su problemi di ottimizzazione del progetto e controllo operativo di Sistemi Energetici Aggregati (in inglese Aggregate Energy Systems) affetti da incertezza parametrica. Questi sistemi sono composti da diverse unità di generazione e stoccaggio di energia, così come fonti non controllabili e distribuite sul territorio. Essi sono progettati con l’obbiettivo di integrare tecnologie tra loro differenti con lo scopo di migliorare l’efficienza complessiva, soddisfare varie domande energetiche e favorire la penetrazione di fonti rinnovabili. Recentemente questi sistemi hanno ricevuto parecchio interesse poiché si sono mostrati competitivi in certe applicazioni (per.es., fornitura di energia termica, elettrica e raffreddamento ad utenze industriali ed urbane) e si prevede che giochino un ruolo di rilievo nel settore energetico del prossimo futuro. In letteratura si possono trovare diverse tipologie di Sistemi Energetici Aggregati: Multi Energy Systems, microgrids, Virtual Power Plants, ecc. I Virtual Power Plants si differenziano dalle altre tipologie di impianto in quanto progettati con lo scopo di partecipare ai mercati elettrici e caratterizzati da una maggiore potenza elettrica installata. Data la presenza di diverse tecnologie e la complessità del sistema, l’ottimizzazione del design ed operation richiede lo sviluppo di strumenti matematici in grado di considerare le particolarità tecniche di ogni unità così come l’incertezza relativa ad alcuni parametri del problema (per.es. produzione da PV). In questa Tesi quattro problemi di interesse industriale sono considerati ed analizzati, e quattro approcci metodologici sono presentati per la loro risoluzione. Questi sono: (1) un approccio di tipo “Rolling Horizon” per l’operation robusta di Sistemi Energetici Aggregati soggetti a vincoli annuali e comprendenti stoccaggi energetici stagionali; (2) un modello “multi-stage” stocastico per l’operation ottima di Virtual Power Plants partecipanti ai mercati elettrici; (3) un modello “two-stage” stocastico per il progetto ottimo di Sistemi Energetici Aggregati soggetti ad affidabilità N-1; (4) un approccio metodologico per il progetto ottimo di un Virtual Power Plant caratterizzato da emissioni nulle di anidride carbonica e stoccaggio stagionale di idrogeno. Gli approcci sviluppati sono stati testati su casi studio di rilevanza industriale risultando efficaci nelle loro funzioni. La schedulazione robusta valutata per mezzo dell’algoritmo Rolling Horizon (1) si è mostrata in grado di servire le domande energetiche dell’utenza quando applicata su di un intero anno e per tutti i casi studio considerati. In aggiunta, i vincoli annuali cui il sistema è soggetto sono stati rispettati e la gestione dell’accumulo stagionale è risultata efficace a prevenire qualsiasi ammanco. Riguardo lo studio sul Virtual Power Plant (2), il modello di ottimizzazione multi-stage stocastico ha dimostrato di essere in grado di modellare in maniera accurata sia i sistemi energetici considerati che i mercati elettrici, valutando soluzioni operative migliori rispetto a quelle ottenute considerando un modello deterministico. In aggiunta, il tempo computazionale è stato ridotto significativamente grazie allo sviluppo di una decomposizione di tipo “Shrinking Horizon”, rendendo l’approccio sviluppato adeguato alla pratica industriale. Per quello che riguarda il terzo argomento di ricerca (3), il modello stocastico “two-stage” sviluppato per la valutazione del design ottimo di Sistemi Energetici Aggregati soggetti ad affidabilità N-1 si è dimostrato capace di valutare soluzioni di impianto in grado di servire le domande delle utenze per la durata di un anno intero ed per qualsiasi evento di rottura. L’elevato tempo computazionale richiesto per la soluzione del modello ha richiesto lo sviluppo di tecniche di decomposizione, di cui una ha mostrato essere molto efficace nell’accorciare i tempi di calcolo sebbene non garantendo l’ottimo globale della soluzione. In fine, l’ultimo tema di ricerca riguarda la valutazione del design ottimo di un possibile futuro Virtual Power Plant comprendente accumulo di idrogeno e cicli combinati basati su tale gas (4). La metodologia combina alcuni aspetti presentati negli approcci sopra citati, come buona parte del modello “two-stage” stocastico del metodo (3), i vincoli di riserva del modello riguardante l’argomento di ricerca (2) e una versione modificata dell’algoritmo Rolling Horizon per l’operation ottima del sistema energetico e della gestione dell’accumulo stagionale riguardante il tema (1). In conclusione, le metodologie presentate in questa Tesi si sono dimostrate in grado di risolvere problemi di rilevanza industriale, sebbene vi sia spazio per miglioramenti riguardanti il tempo computazionale e la facilità di uso ed implementazione da parte dell’utente.
Optimal design and operation of aggregate energy systems under uncertainty
CASTELLI, ALESSANDRO FRANCESCO
2022/2023
Abstract
This thesis focuses on the optimal design and operation of Aggregate Energy Systems under uncertainty. These systems are composed by different dispatchable storage and generation units, as well as non-dispatchable, distributed energy sources. They are designed with the purpose of integrating different energy technologies so to improve the overall efficiency, meet different energy demands and facilitate the penetration of renewable energy sources. Much interest is nowadays present for these systems, since they are already economically competitive in certain applications (e.g., serving heat, power and cooling for industrial and urban users) and expected to play a major role in the energy sector. In literature different kinds of Aggregate Energy Systems can be found: Multi Energy Systems, microgrids, Virtual Power Plants, etc. Virtual Power Plant differentiate themselves from the rest since they are designed with the main goal to participate in the electricity markets, and they usually feature a larger installed power. Given the presence of different technologies and the complexity of the system, both the optimal design and operation require the development of mathematical tools able to capture the different technical aspects of each unit as well as the uncertain nature of the parameters (e.g., PV generation). In this Thesis, four topics of industrial interest are considered and analyzed, and four different approaches are proposed. These are: (1) a Rolling Horizon approach for the optimal robust operation of Aggregate Energy Systems featuring yearly constraints and/or seasonal storage; (2) a multi-stage stochastic optimization model for the optimal operation of Virtual Power Plants participating in the electricity markets; (3) a two-stage stochastic optimization model for the optimal design of Aggregate Energy Systems with (N-1)-reliability; (4) a methodology for the optimal design of zero carbon emission Virtual Power Plants with hydrogen seasonal storage. When applied to real-world case studies, the approaches resulted successful in the tasks they were designed for. At first, the optimal robust operation evaluated by the Rolling Horizon (1) was able to serve the user’s energy demands through an entire year and for all the case studies considered. In addition, the considered yearly constraints were met and the daily management of the seasonal storage proved to be successful in avoiding shortages. Regarding the study on the Virtual Power Plant (2), the multi-stage stochastic optimization model was effective in carefully modelling both the energy systems and the electricity markets, and the solutions obtained proved to be better than the ones obtained with a deterministic approach. Moreover, the computational time was significantly improved by developing a shrinking horizon-based decomposition, making the methodology suitable for practical industrial use. For what concerns the third research topic addressed in this thesis (3), the two-stage stochastic model for the optimal design of Aggregate Energy Systems with (N-1)-reliability was able to compute design solutions able to meet the users’ demands along a whole yearly horizon, for whichever contingency. The high solution time required the development of decomposition techniques, one of them very effective in shortening the computational time although not guaranteeing optimality. Finally, the optimal design of a future Virtual Power Plant featuring hydrogen storage and hydrogen-fired combine cycles was evaluated (4). The proposed methodology combines some of the features presented in the previous works, such as many parts of two-stage stochastic model of (3), the reserve constraints present in (2) and a modified version of the Rolling Horizon for the optimal schedule of the plant and seasonal storage management presented in (1). Results proved these concepts and evaluateed the techno-economic feasibility of such aggregate system. In conclusion, the methodologies presented in this Thesis were proven to be effective for the considered real-world case studies, although improvements can be made in terms of computational time and ease of implementation.| File | Dimensione | Formato | |
|---|---|---|---|
|
THESIS-CASTELLI-final3_revised.pdf
non accessibile
Descrizione: tesi rivista e completata
Dimensione
10.74 MB
Formato
Adobe PDF
|
10.74 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/10589/207586