The energy performance is a relevant matter in the life cycle management of buildings in order to guarantee efficiency, affordability and compliance with the environmental and social purposes for sustainability in the long-term period. Accordingly, buildings’ energy efficiency is planned in the design phase and it is calculated according to procedure stated by Laws; nevertheless, the actual performance of the building differs by the predicted one due to factors associated to the uncertainties diffused in the modelling, construction and operating phases. In predicting the energy performance, design assumption and modelling tools define the boundaries of uncertainty while discussing about real performance built quality, occupancy behavior and management & controls determine the strong variability in the energy results. Therefore, building energy performance simulation requires models, which describe physical phenomena with different levels of detail and accuracy. Detailed dynamic models are accurate but on the other hand require detailed input data and the simulations are time-consuming whereas surrogate models consider only the most relevant parameters that contribute to outline the energy performance. The proposed methodology combines the two model strategies using the detailed simulations to train two Artificial Neural Network (ANN) capable of assessing the heating and cooling demands based on climate and occupancy data. The trained ANNs can predict energy performance of the building with different occupancy rates reducing the use of time-expensive detailed simulations. Moreover, in a Building Management System (BMS) ANNs may be fed by real-time data acquired by sensors and control the settings of systems and devices (e.g. HVAC, shading devices, artificial lighting, etc.). In the paper the Smart Campus Demonstrator or eLUX lab, a university building located in Brescia, Italy, is used as a case study to apply this methodology aiming into identify a range of performance reliability considering the users’ dependent segment of thermal consumption.
Predicting Energy Performance of an Educational Building through Artificial Neural Network
Lavinia Chiara Tagliabue
;Angelo Luigi Camillo Ciribini;Enrico De Angelis
2016-01-01
Abstract
The energy performance is a relevant matter in the life cycle management of buildings in order to guarantee efficiency, affordability and compliance with the environmental and social purposes for sustainability in the long-term period. Accordingly, buildings’ energy efficiency is planned in the design phase and it is calculated according to procedure stated by Laws; nevertheless, the actual performance of the building differs by the predicted one due to factors associated to the uncertainties diffused in the modelling, construction and operating phases. In predicting the energy performance, design assumption and modelling tools define the boundaries of uncertainty while discussing about real performance built quality, occupancy behavior and management & controls determine the strong variability in the energy results. Therefore, building energy performance simulation requires models, which describe physical phenomena with different levels of detail and accuracy. Detailed dynamic models are accurate but on the other hand require detailed input data and the simulations are time-consuming whereas surrogate models consider only the most relevant parameters that contribute to outline the energy performance. The proposed methodology combines the two model strategies using the detailed simulations to train two Artificial Neural Network (ANN) capable of assessing the heating and cooling demands based on climate and occupancy data. The trained ANNs can predict energy performance of the building with different occupancy rates reducing the use of time-expensive detailed simulations. Moreover, in a Building Management System (BMS) ANNs may be fed by real-time data acquired by sensors and control the settings of systems and devices (e.g. HVAC, shading devices, artificial lighting, etc.). In the paper the Smart Campus Demonstrator or eLUX lab, a university building located in Brescia, Italy, is used as a case study to apply this methodology aiming into identify a range of performance reliability considering the users’ dependent segment of thermal consumption.File | Dimensione | Formato | |
---|---|---|---|
Predicting Energy Performance of an Educational Building through Artificial Neural Network.pdf
gestori archivio
Tipologia:
Full Text
Licenza:
NON PUBBLICO - Accesso privato/ristretto
Dimensione
958.29 kB
Formato
Adobe PDF
|
958.29 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.