logo SBA

ETD

Digital archive of theses discussed at the University of Pisa

 

Thesis etd-09202022-080912


Thesis type
Tesi di laurea magistrale
Author
BIZZARRI, MARCO
URN
etd-09202022-080912
Thesis title
Building heating load prediction for optimal control of hybrid heat pumps: evaluation of an autoregressive model and energy analysis
Department
INGEGNERIA DELL'ENERGIA, DEI SISTEMI, DEL TERRITORIO E DELLE COSTRUZIONI
Course of study
INGEGNERIA ENERGETICA
Supervisors
relatore Prof. Testi, Daniele
relatore Dott.ssa Schito, Eva
relatore Dott. Conti, Paolo
Keywords
  • optimal control
  • autoregressive model
  • heating load forecasting
  • hybrid heat pump
Graduation session start date
06/10/2022
Availability
Withheld
Release date
06/10/2092
Summary
The present work aims to develop a control strategy to seek the optimal operation of a hybrid system, constituted by an electrically driven heat pump and a condensing boiler, by selecting the most suitable supply temperature to be provided to terminal units and choosing the most appropriate thermal generator to be activated. 
To do so, the performances of the two generators, mainly depending on partial load conditions and supply temperature, need to be assessed in advance. Therefore, a short-term prediction of the heating demand is obtained by means of an autoregressive model, which provides an estimate of the load only monitoring the energy requests in the previous hours, after a training period. A regression model is also used to forecast the optimal supply temperature as a function of the heating load. Those predictions are used within the generator models to assess their performance and subsequently choose the most suitable configuration to meet the heating demand.
The predictive accuracy of the autoregressive model, resulting from the application to the most frequent climatic conditions in Italy, is acceptable for non-modern building types, giving average errors on load prediction in the 6-12% range, which, however, can worsen significantly if indoor occupancy profiles are more irregular. Nevertheless, even in those cases, by applying the control strategy to the hybrid system, savings up to 20% were estimated, compared to the control strategy currently used in commercial applications.
Conversely, in more recent buildings, the classical control strategy is not improved, since the relevance of highly variable internal and solar inputs causes the prediction to deteriorate significantly, compromising the control optimality.
Finally, the forecasting method was also tested on measurement data, obtained on real building heating systems, leading to an early validation of the prediction model.
File