Within the framework of EU’s strategy to actuate an effective and sustainable energy transition, scientific research efforts are targeted at cutting greenhouse gases emissions and promoting energy efficiency through multidisciplinary interventions. The context of a smart city manages to host an ensemble of latest generation technologies in terms of measuring devices, informative infrastructures and energy management systems which perfectly complies with this purpose. To further improve their performances, energy fluxes within a smart city can be optimised through the evolution towards a Renewable Energy Community. Moreover, the latter can be supported by a data-driven design approach, in such a way to adapt its management to the energy consumption habits of its members. To achieve this objective, raw data are elaborated with two distinct methodologies, and fed to a Machine Learning algorithm: K-Means. Hence, the present study aims at observing real electricity withdrawal profiles measured by smart meters installed within REDO, a technologically-advanced district dedicated to social housing, with the goal of recognizing typical consumption behaviours among its inhabitants. In other words, daily consumption profiles of households have been clustered based on their similarity in terms of shape and vertical proximity. After associating each daily profile to a set of characteristics, a procedure is designed to determine the optimal number of clusters. Such strategy consists in a semi-empirical analysis, implemented on each produced group of data, in terms of features uniformity, features cardinality and centroids shape, assisted by analytical methods such as Silhouette score and Elbow Method. Consequently, once the optimal number of groups is defined, each of them is characterised based on the prevailing features that it contains. Finally, in view of aggregating smart cities households into a Renewable Energy Community, a set of KPI is identified to evaluate its performances from environmental, economic and technical standpoints, in order to verify that the corresponding goals of the community are achieved.
Nel quadro della strategia dell’Unione Europea per attuare una transizione energetica efficace e sostenibile, gli sforzi della ricerca scientifica mirano a ridurre le emissioni di gas serra e promuovere l'efficienza energetica attraverso interventi multidisciplinari. Il contesto di una smart city riesce ad ospitare un insieme di tecnologie di ultima generazione in termini di strumenti di misura, infrastrutture informative e sistemi di gestione dell'energia che risponde perfettamente a tale scopo. Per migliorare ulteriormente le loro prestazioni, i flussi energetici all'interno di una città intelligente possono essere ottimizzati attraverso il passaggio a una Comunità Energetica Rinnovabile. Quest'ultima, inoltre, può essere supportata da un dimensionamento basato sui dati, in modo tale da adattarne la successiva gestione alle abitudini di consumo energetico dei suoi membri. Per raggiungere questo obiettivo, i dati grezzi vengono elaborati con due metodologie distinte e forniti a un algoritmo di Machine Learning: K-Means. Pertanto, questo studio si propone di osservare i profili reali di prelievo di energia elettrica misurati da smart meters installati all'interno di REDO, un distretto tecnologicamente avanzato dedicato al social housing, con l'obiettivo di riconoscere comportamenti di consumo tipici dei suoi abitanti. In altre parole, i profili di consumo giornaliero delle famiglie sono stati raggruppati in base alla loro somiglianza in termini di forma e prossimità verticale. Dopo aver associato ogni profilo giornaliero a un insieme di caratteristiche, viene progettata una procedura per determinare il numero ottimale di cluster. Tale strategia consiste in un'analisi semi-empirica, attuata su ciascun gruppo di dati creato, in termini di uniformità delle caratteristiche, cardinalità delle caratteristiche e forma dei centroidi, assistita da metodi analitici come Silhouette Score e Elbow Method. Di conseguenza, una volta definito il numero ottimale di gruppi, ciascuno di essi viene caratterizzato in base alle caratteristiche prevalenti che contiene al suo interno. Infine, nell'ottica dell'aggregazione delle famiglie delle smart city in una Renewable Energy Community, viene individuato un set di KPI per valutarne le performance dal punto di vista ambientale, economico e tecnico e verificare, quindi, il corretto raggiungimento degli obiettivi corrispondenti.
Clustering of residential electricity consumption profiles and KPIs identification for Renewable Energy Communities
Siboni, Lorenzo
2021/2022
Abstract
Within the framework of EU’s strategy to actuate an effective and sustainable energy transition, scientific research efforts are targeted at cutting greenhouse gases emissions and promoting energy efficiency through multidisciplinary interventions. The context of a smart city manages to host an ensemble of latest generation technologies in terms of measuring devices, informative infrastructures and energy management systems which perfectly complies with this purpose. To further improve their performances, energy fluxes within a smart city can be optimised through the evolution towards a Renewable Energy Community. Moreover, the latter can be supported by a data-driven design approach, in such a way to adapt its management to the energy consumption habits of its members. To achieve this objective, raw data are elaborated with two distinct methodologies, and fed to a Machine Learning algorithm: K-Means. Hence, the present study aims at observing real electricity withdrawal profiles measured by smart meters installed within REDO, a technologically-advanced district dedicated to social housing, with the goal of recognizing typical consumption behaviours among its inhabitants. In other words, daily consumption profiles of households have been clustered based on their similarity in terms of shape and vertical proximity. After associating each daily profile to a set of characteristics, a procedure is designed to determine the optimal number of clusters. Such strategy consists in a semi-empirical analysis, implemented on each produced group of data, in terms of features uniformity, features cardinality and centroids shape, assisted by analytical methods such as Silhouette score and Elbow Method. Consequently, once the optimal number of groups is defined, each of them is characterised based on the prevailing features that it contains. Finally, in view of aggregating smart cities households into a Renewable Energy Community, a set of KPI is identified to evaluate its performances from environmental, economic and technical standpoints, in order to verify that the corresponding goals of the community are achieved.File | Dimensione | Formato | |
---|---|---|---|
2022_9_Siboni.pdf
accessibile in internet per tutti
Descrizione: Clustering of residential electricity consumption profiles and KPIs identification for Renewable Energy Communities
Dimensione
2.47 MB
Formato
Adobe PDF
|
2.47 MB | Adobe PDF | Visualizza/Apri |
2022_9_Siboni_Extended_Abstract.pdf
accessibile in internet per tutti
Descrizione: Extended Abstract - Clustering of residential electricity consumption profiles and KPIs identification for Renewable Energy Communities
Dimensione
858.77 kB
Formato
Adobe PDF
|
858.77 kB | 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/10589/195652