In this thesis we discuss performances, advantages and drawbacks in the application of two different schemes in the specific context of the energy management of a district composed of multiple buildings sharing a thermal storage. The problem experimental set up is structured so as the multiple buildings cooperate eachother in order to minimize their cooling energy cost: to this purpose they act on their own local decision variables-i.e. the buildings temperature set-point and the usage of the common device-can be set while guaranteeing comfort conditions for their occupants and to properly manage the shared thermal storage. Buildings decision variables are coupled through the constraints on the thermal storage and through the cost function being the electric energy price depending on the district demand. The first algorithm had been recently proposed in literature and consists in a nested optimization program where an ”outer layer” deals with global constraints by updating some dual variables whereas the primal variables are optimized locally in a decentralized fashion within an iterative ”inner layer”. The second algorithm is well known and established within the scientific community and finds wide application in a number of areas, especially statistics and machine learning. It allows to split optimization problems into local and global subproblems accounting for separable structure of the objective function. The comparison between these two methods consists in analyzing differences in the values of the energy price and in graphical outcomes related to the behavior of the optimization variables and the quantities that depend upon them. A second analysis will be provided pointing out the reasons behind these differences, which can be addressed to the different computing times to reach convergence by the algorithms we are dealing with.
Lo scopo di questa tesi consiste nel discutere le prestazioni, i vantaggi e gli svantaggi dell’applicazione di due diversi algoritmi iterativi nel contesto specifico della gestione ottima energetica in una rete di palazzi che condividono un serbatoio in comune. Il set up sperimentale del problema viene strutturato in modo che i palazzi collaborino tra di loro in modo da minimizzare il costo di distribuzione dell’energia di raffreddamento all’interno della rete: a tal proposito agiscono sulle loro variabili decisionali locali-quali il set-point delle temperature e l’utilizzo della risorsa comune-in modo da garantire le condizioni di comfort per gli occupanti e per gestire appropriatamente il serbatoio condiviso. Le variabile decisionali dei palazzi sono accoppiate tramite vincoli globali sulla risorsa in comune e tramite il prezzo dell’energia in quanto quest’ultima viene espressa in funzione della richiesta dell’intera rete. Il primo algoritmo stato da poco introdotto e consiste in un problema di ottimizzazione vincolata che consta di due cicli annidati strutturati come segue: lo ”strato esterno” relativo ai vincoli globali sull’utilizzo del serbatoio in comune e che si occupa di aggiornare variabili duali, mentre nello ”strato interno” vengono calcolate le variabili decisionali mediante un problema di ottimizzazione locale a livello dei palazzi. Il secondo algoritmo invece ben noto alla comunit scientifica e trova molteplici applicazione in diverse aree di ricerca, quali statistica e machine learning: permette di suddividere il problema principale in sottoproblemi a livello locale (palazzi) coordinato a un problema di ottimizzazione a livello globale. Il paragone tra questi due algoritmi consiste nell’analizzare le differenze tra i valori del prezzo dell’enrgia e le differenze relative alle grandezze espresse in funzione delle variabili di ottimizzazione. Una seconda analisi viene dedicata ai tempi di calcolo grazie a cui i metodi in esame raggiungono la convergenza.
Two different approaches to the optimal management of a district network with a shared thermal storage
DE BELLA, PAOLO
2016/2017
Abstract
In this thesis we discuss performances, advantages and drawbacks in the application of two different schemes in the specific context of the energy management of a district composed of multiple buildings sharing a thermal storage. The problem experimental set up is structured so as the multiple buildings cooperate eachother in order to minimize their cooling energy cost: to this purpose they act on their own local decision variables-i.e. the buildings temperature set-point and the usage of the common device-can be set while guaranteeing comfort conditions for their occupants and to properly manage the shared thermal storage. Buildings decision variables are coupled through the constraints on the thermal storage and through the cost function being the electric energy price depending on the district demand. The first algorithm had been recently proposed in literature and consists in a nested optimization program where an ”outer layer” deals with global constraints by updating some dual variables whereas the primal variables are optimized locally in a decentralized fashion within an iterative ”inner layer”. The second algorithm is well known and established within the scientific community and finds wide application in a number of areas, especially statistics and machine learning. It allows to split optimization problems into local and global subproblems accounting for separable structure of the objective function. The comparison between these two methods consists in analyzing differences in the values of the energy price and in graphical outcomes related to the behavior of the optimization variables and the quantities that depend upon them. A second analysis will be provided pointing out the reasons behind these differences, which can be addressed to the different computing times to reach convergence by the algorithms we are dealing with.File | Dimensione | Formato | |
---|---|---|---|
2018_04_DeBella.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
677.78 kB
Formato
Adobe PDF
|
677.78 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/10589/141078