This thesis addresses the optimal design of urban energy districts, also called Multi-Energy Systems (MESs) or District Energy Systems (DESs), wherein multiple energy demands must be satisfied. The MES/DES might feature multiple locations with users, energy conversion and storage systems. Inner networks allow the distribution of energy, from the conversion and storage units to the users, and the exchange between all of them. The design optimisation of a MES/DES can not prescind from its operation in the various conditions the it would face throughout the year. Yet, this way, the problem would become intractable. To reduce the size of the problem, clustering algorithms are used to select a number of typical and extreme days to be considered for the operation. A novel clustering algorithm is proposed and tested against traditional ones on the design optimisation of the MES of two case studies. It turns out to be a valid alternative to well-known clustering techniques for its accuracy in reproducing both the hourly profiles and the integral of the aggregated load duration curves, while automatically identifying extreme and a-typical operating periods. The novelty of the optimisation model proposed in this thesis is twofold: (i) the district can have a bidirectional exchange of electricity with the national grid, entailing the participation to a two-stage electricity market (day-ahead bidding and real-time balancing); (ii) the heat integration of the thermal units can be both in parallel and series configurations, so that district heating networks with large delivery-return temperature difference can be properly modelled. To address the first aspect, I propose a multi-scale three-stage stochastic mixed-integer linear programming model (SMILP) with integer recourse that simultaneously optimises the investment and the operational decisions. The operational decisions comprise both day-ahead unit commitment of the slow start-up energy conversion systems (e.g., biomass-fired organic Rankine cycles) and real-time scheduling adjustment (integer recourse) of the fast start-up ones (e.g., heat pumps). The model explicitly considers the impact of the errors (i.e., the uncertainty) in the day-ahead forecasts of the energy demand profiles and the production from non-dispatchable renewable energy sources on the optimal design of a MES. In order to overcome the computational challenge posed by the three-stage SMILP problem, I propose a heuristic algorithm which solves a sequence of design and operation problems with progressively increasing integrality restrictions. I apply the proposed modelling framework to a university campus and demonstrate the advantages with respect to the deterministic approach. In light of the results obtained by the stochastic approach, the benefits of MESs optimisation are further investigated following a deterministic approach (disregarding the uncertainty associated to the day-ahead forecasts) and applying the optimisation model to three case studies, which show that the optimal selection and sizing of technologies is case-specific. The multiple heat integration arrangements has required the development of a temperature intervals superstructure, thanks to which the optimisation model can properly account for the heat contributions from the units, according to their inlet and outlet temperature. The optimisation model has been used to compare a number of retrofit options for the heating and electricity supply to a large-size energy district.
Questa tesi affronta la progettazione ottimale di distretti energetici urbani, chiamati anche Multi-Energy Systems (MES) o District Energy Systems (DES), in cui più domande energetiche devono essere soddisfatte. Il MES/DES potrebbe presentare più siti di produzione, stoccaggio e consumo dell’energia. Le reti interne consentono lo scambio di energia tra siti. L'ottimizzazione del progetto di un MES/DES non può prescindere dall’operation (funzionamento ora per ora) nelle varie condizioni che dovrebbe affrontare durante l'anno. Tuttavia, in questo modo, il problema diventerebbe computazionalmente intrattabile. Per ridurre le dimensioni del problema, vengono utilizzati algoritmi di clustering per selezionare un i giorni tipo ed estremi da considerare nella risoluzione. Un nuovo algoritmo di clustering è proposto e testato. Esso risulta essere una valida alternativa agli approcci più noti di clustering, grazie alla sua precisione nel riprodurre sia i profili orari sia l'integrale delle curve di carico, identificando automaticamente periodi operativi estremi e tipici. La novità del modello di ottimizzazione proposto in questa tesi è duplice: (i) il distretto può avere uno scambio bidirezionale di elettricità con la rete nazionale, implicando la partecipazione a un mercato dell'elettricità a due livelli (offerta del giorno prima e bilanciamento in tempo reale); (ii) l'integrazione termica delle unità può essere sia in parallela che in serie, in modo che le reti di teleriscaldamento, caratterizzate da un’elevata differenza di temperatura tra mandata e ritorno, possano essere adeguatamente modellate. Per affrontare il primo aspetto, viene proposto un problema ‘’three-stage stochastic mixed-integer linear programming’’ (SMILP) con ricorso intero che ottimizza simultaneamente le decisioni di investimento e operative. Le decisioni operative comprendono sia lo unit commitment del giorno prima dei sistemi di conversione dell'energia ad avvio lento (ad es., Cicli Rankine organici alimentati a biomassa) sia l'adeguamento del programma di gestione in tempo reale (ricorso intero) di quelli ad avvio rapido (ad es., pompe di calore). Il modello considera esplicitamente l'impatto degli errori (ossia l’incertezza) nelle previsioni del giorno prima dei profili di domanda di energia e della produzione da fonti di energia rinnovabile non-dispacciabili sulla progettazione ottimale del distretto. Il problema stocastico così formulato risulta computazionalmente intrattabile, quindi, al fine di individuare una approssimazione della soluzione ottimale, viene proposto un algoritmo euristico che risolve una sequenza di problemi di progettazione e operation con l’aggiunta progressiva di vincoli integrali. Il modello stocastico proposto viene applicato a un campus universitario e si dimostrano i vantaggi rispetto all'approccio deterministico. I vantaggi dell'ottimizzazione dei MES sono ulteriormente studiati seguendo un approccio deterministico (trascurando l'incertezza associata alle previsioni del giorno prima) considerando tre casi studio, che dimostrano che la selezione ottimale e il dimensionamento dei sistemi di conversione e stoccaggio dell’energia sono caso-specifici. La modellazione di configurazioni di integrazione termica serie/parallelod hanno richiesto lo sviluppo di una sovrastruttura a intervalli di temperatura, grazie alla quale il modello di ottimizzazione può tenere adeguatamente in considerazione i contributi termici specifici delle singole unità, in base alla loro temperatura di ingresso e uscita. Il modello di ottimizzazione è stato utilizzato per confrontare una serie di opzioni di retrofit per il riscaldamento e la fornitura di elettricità a un distretto energetico di grandi dimensioni.
Optimal design of urban energy districts under uncertainty
ZATTI, MATTEO
Abstract
This thesis addresses the optimal design of urban energy districts, also called Multi-Energy Systems (MESs) or District Energy Systems (DESs), wherein multiple energy demands must be satisfied. The MES/DES might feature multiple locations with users, energy conversion and storage systems. Inner networks allow the distribution of energy, from the conversion and storage units to the users, and the exchange between all of them. The design optimisation of a MES/DES can not prescind from its operation in the various conditions the it would face throughout the year. Yet, this way, the problem would become intractable. To reduce the size of the problem, clustering algorithms are used to select a number of typical and extreme days to be considered for the operation. A novel clustering algorithm is proposed and tested against traditional ones on the design optimisation of the MES of two case studies. It turns out to be a valid alternative to well-known clustering techniques for its accuracy in reproducing both the hourly profiles and the integral of the aggregated load duration curves, while automatically identifying extreme and a-typical operating periods. The novelty of the optimisation model proposed in this thesis is twofold: (i) the district can have a bidirectional exchange of electricity with the national grid, entailing the participation to a two-stage electricity market (day-ahead bidding and real-time balancing); (ii) the heat integration of the thermal units can be both in parallel and series configurations, so that district heating networks with large delivery-return temperature difference can be properly modelled. To address the first aspect, I propose a multi-scale three-stage stochastic mixed-integer linear programming model (SMILP) with integer recourse that simultaneously optimises the investment and the operational decisions. The operational decisions comprise both day-ahead unit commitment of the slow start-up energy conversion systems (e.g., biomass-fired organic Rankine cycles) and real-time scheduling adjustment (integer recourse) of the fast start-up ones (e.g., heat pumps). The model explicitly considers the impact of the errors (i.e., the uncertainty) in the day-ahead forecasts of the energy demand profiles and the production from non-dispatchable renewable energy sources on the optimal design of a MES. In order to overcome the computational challenge posed by the three-stage SMILP problem, I propose a heuristic algorithm which solves a sequence of design and operation problems with progressively increasing integrality restrictions. I apply the proposed modelling framework to a university campus and demonstrate the advantages with respect to the deterministic approach. In light of the results obtained by the stochastic approach, the benefits of MESs optimisation are further investigated following a deterministic approach (disregarding the uncertainty associated to the day-ahead forecasts) and applying the optimisation model to three case studies, which show that the optimal selection and sizing of technologies is case-specific. The multiple heat integration arrangements has required the development of a temperature intervals superstructure, thanks to which the optimisation model can properly account for the heat contributions from the units, according to their inlet and outlet temperature. The optimisation model has been used to compare a number of retrofit options for the heating and electricity supply to a large-size energy district.File | Dimensione | Formato | |
---|---|---|---|
2019_09_PhD_Zatti.pdf
accessibile in internet per tutti
Descrizione: Thesis text
Dimensione
7.27 MB
Formato
Adobe PDF
|
7.27 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/10589/148435