The urgent need to combat climate change highlights the critical importance of transitioning to renewable energy and achieving net-zero emissions by 2050. Key strategies include electrifying the transportation sector through electric vehicle (EV) adoption, expanding solar photovoltaic (PV) capacity, and integrating renewable energy sources (RES) into distributed energy systems. These advancements, however, pose significant challenges, such as ensuring grid stability and effectively coordinating RES generation with EV consumption. Digitalization and advanced optimization algorithms are essential for addressing these complexities. Energy Management Systems (EMS), particularly in residential contexts, optimize energy use, reduce costs, and support high RES penetration while balancing user needs and managing system uncertainties. This thesis focuses on the development of hierarchical EMS to optimize the operation of microgrids and smart homes, with an emphasis on integrating RES and EV demands. By adopting predictive deterministic and stochastic optimization techniques, the work overcomes the limitations of existing heuristic approaches. Using Mixed Integer Linear Programming (MILP) for optimal scheduling, the thesis addresses uncertainties in energy production, residential demand, and EV usage. Extensive real-world case studies conducted in collaboration with industrial partners demonstrate the advantages of the proposed predictive EMS, achieving annual cost savings exceeding 20% and reaching up to €900/year for configurations with large PV systems, with additional benefits when vehicle-to-home (V2H) functionality is enabled. The analysis identifies EV usage as the primary source of uncertainty, leading to the stochastic enhancement of the deterministic home EMS. Developed in partnership with KU Leuven’s Department of Electrical Engineering, the stochastic EMS incorporates probabilistic forecasts of PV generation, residential consumption, and EV availability. The work emphasizes the trade-off between accurately modeling uncertainty and the computational burden of optimization. A two-stage stochastic MILP framework is proposed, compared across formulations, and validated, demonstrating consistent resilience and superior performance over deterministic methods. Moreover, the stochastic EMS is shown to be deployable in practical applications. Experimental validations are conducted using the Multi-Good MicroGrid Laboratory (MG2Lab) and real-world smart homes in collaboration with Edison. These activities assess the impact of forecast accuracy on operational costs and validate the EMS software with real customers, confirming its effectiveness, resilience, and real-world applicability.
L’urgenza di contrastare il cambiamento climatico rende fondamentale la transizione verso l’energia rinnovabile e il raggiungimento delle emissioni nette zero entro il 2050. Tra le strategie chiave emergono l’elettrificazione del settore dei trasporti tramite l’adozione di veicoli elettrici (EV), l’espansione della capacità fotovoltaica (PV) e l’integrazione delle fonti rinnovabili (RES) nei sistemi energetici distribuiti. Tuttavia, questi progressi comportano sfide significative, come il mantenimento della stabilità della rete e il coordinamento efficace tra la generazione da fonti rinnovabili e il consumo dei veicoli elettrici. In questo contesto, la digitalizzazione e l’uso di algoritmi di ottimizzazione avanzati diventano essenziali per affrontare tali complessità. I sistemi di gestione dell’energia (EMS), in particolare nei contesti residenziali, svolgono un ruolo cruciale nell’ottimizzazione dell’uso energetico, nella riduzione dei costi e nel supporto a un’elevata penetrazione delle energie rinnovabili, garantendo al contempo il rispetto delle esigenze degli utenti e la gestione delle incertezze di sistema. Questa tesi si concentra sullo sviluppo di EMS gerarchici per ottimizzare il funzionamento delle microreti e delle smart home, con un’attenzione particolare all’integrazione della generazione rinnovabile e della domanda degli EV. Attraverso l’adozione di tecniche di ottimizzazione predittiva, sia deterministica che stocastica, il lavoro supera i limiti degli approcci euristici esistenti. L’uso della programmazione lineare intera mista (MILP) per la pianificazione ottimale consente di affrontare le incertezze legate alla produzione energetica, alla domanda residenziale e all’uso dei veicoli elettrici. I numerosi casi di studio realizzati in collaborazione con partner industriali dimostrano i vantaggi degli EMS predittivi proposti, con risparmi annui superiori al 20% e fino a 900 €/anno in configurazioni con grandi impianti fotovoltaici, con ulteriori benefici quando è abilitata la funzionalità di vehicle-to-home (V2H). L’analisi individua l’uso degli EV come la principale fonte di incertezza, portando all’evoluzione stocastica dell’EMS domestico deterministico. In collaborazione con il Dipartimento di Ingegneria Elettrica della KU Leuven, è stato sviluppato un EMS stocastico che incorpora previsioni probabilistiche della generazione PV, del consumo residenziale e della disponibilità dei veicoli elettrici. Il lavoro pone l’accento sul bilanciamento tra l’accuratezza nella modellazione dell’incertezza e l’onere computazionale dell’ottimizzazione. Viene proposto un framework stocastico MILP a due stadi, confrontato tra diverse formulazioni e validato, dimostrando una resilienza costante e prestazioni superiori rispetto ai metodi deterministici. Inoltre, si dimostra che l’EMS stocastico è concretamente applicabile in contesti reali. La validazione sperimentale è stata effettuata utilizzando il Multi-Good MicroGrid Laboratory (MG2Lab) e smart home reali in collaborazione con Edison. Queste attività hanno permesso di valutare l’impatto dell’accuratezza delle previsioni sui costi operativi e di verificare l’efficacia del software EMS con utenti reali, confermandone l’efficienza, la resilienza e l’applicabilità pratica.
Optimization and control of microgrids with high renewables penetration for the supply of hybrid and electric vehicles
GIOFFRÈ, DOMENICO
2024/2025
Abstract
The urgent need to combat climate change highlights the critical importance of transitioning to renewable energy and achieving net-zero emissions by 2050. Key strategies include electrifying the transportation sector through electric vehicle (EV) adoption, expanding solar photovoltaic (PV) capacity, and integrating renewable energy sources (RES) into distributed energy systems. These advancements, however, pose significant challenges, such as ensuring grid stability and effectively coordinating RES generation with EV consumption. Digitalization and advanced optimization algorithms are essential for addressing these complexities. Energy Management Systems (EMS), particularly in residential contexts, optimize energy use, reduce costs, and support high RES penetration while balancing user needs and managing system uncertainties. This thesis focuses on the development of hierarchical EMS to optimize the operation of microgrids and smart homes, with an emphasis on integrating RES and EV demands. By adopting predictive deterministic and stochastic optimization techniques, the work overcomes the limitations of existing heuristic approaches. Using Mixed Integer Linear Programming (MILP) for optimal scheduling, the thesis addresses uncertainties in energy production, residential demand, and EV usage. Extensive real-world case studies conducted in collaboration with industrial partners demonstrate the advantages of the proposed predictive EMS, achieving annual cost savings exceeding 20% and reaching up to €900/year for configurations with large PV systems, with additional benefits when vehicle-to-home (V2H) functionality is enabled. The analysis identifies EV usage as the primary source of uncertainty, leading to the stochastic enhancement of the deterministic home EMS. Developed in partnership with KU Leuven’s Department of Electrical Engineering, the stochastic EMS incorporates probabilistic forecasts of PV generation, residential consumption, and EV availability. The work emphasizes the trade-off between accurately modeling uncertainty and the computational burden of optimization. A two-stage stochastic MILP framework is proposed, compared across formulations, and validated, demonstrating consistent resilience and superior performance over deterministic methods. Moreover, the stochastic EMS is shown to be deployable in practical applications. Experimental validations are conducted using the Multi-Good MicroGrid Laboratory (MG2Lab) and real-world smart homes in collaboration with Edison. These activities assess the impact of forecast accuracy on operational costs and validate the EMS software with real customers, confirming its effectiveness, resilience, and real-world applicability.File | Dimensione | Formato | |
---|---|---|---|
Gioffrè_Phd_Thesis.pdf
non accessibile
Dimensione
13.29 MB
Formato
Adobe PDF
|
13.29 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/10589/237072