The purpose of this thesis consists in the evaluation of the accuracy of a data-driven model, that was built to predict the behaviour of two geothermal residential heat pumps with the same components, and one of them is characterized by the presence of a desuperheater. The analysis of the heat pumps dynamics is based on the use of on-field measurements, that provide a picture of the actual behaviour of the systems under investigation. Different studies rely on real measurements to monitor HVAC systems, but only a tiny share of them is focused on residential application, this thesis is proposed as a piece to cover this gap in this research topic. The model was developed using Python as programming language. The methodology adopted to pursue the objective of this thesis is made of two main steps. First, data provided by the manufacturer were carefully analysed and pre processed to have a suitable dataset. Second, the model has been modified and tested and its accuracy has been evaluated through the use of significant statistical parameters; i.e., R 2 for all variables of interest, Relative Root Mean Square Error (RMSE) for temperatures and Relative RMSE (RRMSE) for pressures and for the compressor work. The results obtained after the resting phase showed that the model is less accurate then expected. The worst result is for evaporation temperature and pressure in both heat pumps both in terms of R2 and of RMSE. The accuracy in the prediction of the other variables proved to be better than evaporation temperature and pressure, even if not perfect. Then, a sensitivity analysis was carried on increasing the resample time of the data. This operation did not lead to a well-defined result, in fact some parameters worsen and others improve, this depends on the dataset under consideration and the heat pump under observation at that time. As highlighted in more detail below, the negative results are strongly influenced by poor data sets provided by the manufacturer. The model tested in this thesis has been shown to be extremely accurate with the use of good quality input data. In order to improve the model, so that it can be used on different heat pumps, it is therefore necessary that higher quality data be provided that more accurately represents the behavior of the system.
Lo scopo di questa tesi consiste nel valutare l’accuratezza di un modello data-driven, realizzato per prevedere il comportamento di due pompe di calore geotermiche residen ziali. Le due pompe sono caratterizzate dagli gli stessi componenti, ma una presenta un desurriscaldatore. I dati forniti dal produttore sono stati raccolti tramite l’ausilio di sensori. Il monitoraggio dei sistemi HVAC viene spesso effettuato con l’utilizzo di misure reali (i.e., non proveniente da prove svolte in ambiente controllato), ma solo una minima parte di essi si concentra su applicazioni residenziali; questa tesi si inserisce in questo gap presente in letteratura. Il modello è stato sviluppato utilizzando il linguaggio di pro grammazione Python. La metodologia adottata per perseguire l’obiettivo di questa tesi è costituita da due fasi principali. In primo luogo, i dati forniti dal produttore sono stati attentamente analizzati e pre-elaborati per ottenere un set di dati adeguato. In seguito, il modello è stato modificato e testato e la sua accuratezza è stata valutata attraverso l’uso di parametri statistici significativi; ovvero, R2 per tutte le variabili di interesse, la deviazione quadratica media (RMSE) per le temperature e la deviazione quadratica media relativa alla media dei dati forniti (RRMSE) per le pressioni e per il lavoro del compres sore. I risultati ottenuti in seguito al test hanno mostrato che il modello è meno accurato del previsto. Il risultato peggiore riguarda la temperatura e la pressione di evaporazione in entrambe le pompe di calore, sia in termini di R2 che di RMSE. L’accuratezza nella previsione delle altre variabili si è dimostrata migliore, anche se non perfetta. È stata poi effettuata un’analisi di sensitività aumentando il tempo di ricampionamento dei dati. Questa operazione non ha portato a un netto, infatti alcuni parametri peggiorano e altri migliorano, ciò dipende dal set di dati in esame e dalla pompa di calore in osservazione in quel momento. Come evidenziato più dettagliatamente in seguito, i risultati negativi sono fortemente influenzati dai set di dati scadenti forniti dal produttore. Il modello su cui si focalizza questa tesi era stato precedentemente testato con l’utilizzo di dati di maggiore qualità, portando ad ottimi risultati. La possibilità di applicare tale modello ad altre pompe di calore ed avere risultati affidabili, dipenderà in maniera consistente dall’utilizzo di dati migliori, che rappresentino in modo più accurato il comportamento del sistema.
Comprehensive experimental validation and performance analysis of a data-driven model for two brine-to-water heat pumps
Dolce, Giovanni
2022/2023
Abstract
The purpose of this thesis consists in the evaluation of the accuracy of a data-driven model, that was built to predict the behaviour of two geothermal residential heat pumps with the same components, and one of them is characterized by the presence of a desuperheater. The analysis of the heat pumps dynamics is based on the use of on-field measurements, that provide a picture of the actual behaviour of the systems under investigation. Different studies rely on real measurements to monitor HVAC systems, but only a tiny share of them is focused on residential application, this thesis is proposed as a piece to cover this gap in this research topic. The model was developed using Python as programming language. The methodology adopted to pursue the objective of this thesis is made of two main steps. First, data provided by the manufacturer were carefully analysed and pre processed to have a suitable dataset. Second, the model has been modified and tested and its accuracy has been evaluated through the use of significant statistical parameters; i.e., R 2 for all variables of interest, Relative Root Mean Square Error (RMSE) for temperatures and Relative RMSE (RRMSE) for pressures and for the compressor work. The results obtained after the resting phase showed that the model is less accurate then expected. The worst result is for evaporation temperature and pressure in both heat pumps both in terms of R2 and of RMSE. The accuracy in the prediction of the other variables proved to be better than evaporation temperature and pressure, even if not perfect. Then, a sensitivity analysis was carried on increasing the resample time of the data. This operation did not lead to a well-defined result, in fact some parameters worsen and others improve, this depends on the dataset under consideration and the heat pump under observation at that time. As highlighted in more detail below, the negative results are strongly influenced by poor data sets provided by the manufacturer. The model tested in this thesis has been shown to be extremely accurate with the use of good quality input data. In order to improve the model, so that it can be used on different heat pumps, it is therefore necessary that higher quality data be provided that more accurately represents the behavior of the system.File | Dimensione | Formato | |
---|---|---|---|
2023_12_Dolce.pdf
accessibile in internet per tutti
Dimensione
6.21 MB
Formato
Adobe PDF
|
6.21 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/10589/215507