Prediction of HVAC System Parameters Using Deep Learning

Sirine Maalej (1) , Zoubeir Lafhaj (2) , Jean Yim , Pascal Yim (3) , Colin Noort (1)

1 BCIT - British Columbia Institute of Technology [Burnaby, BC]
2 Centrale Lille
3 PERSI - PERennisation des Systèmes Industriels

Heating, ventilation, and air conditioning (HVAC) systems consume between 10-20% of developed countries' energy annually. Up to 30% of this energy is often wasted due to mismanagement or improper control strategies. In order to overcome this issue and optimize energy consumption, this paper proposes a predictive modeling technique to effectively forecast HVAC system parameters using machine and deep learning models. A case study of an air handling unit (AHU) at the British Columbia Institute of Technology (BCIT, Canada) is used to test and confirm the results of this research. Five models were applied to predict the supply air temperature. Each model was compared with actual supply air temperature and its accuracy was explored. The results reveal that all investigated models were successful in predicting the supply air temperature, and that the combination of a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) models has obtained the highest accuracy.

Domaines: Apprentissage [cs.LG]; Construction durable

Sirine Maalej, Zoubeir Lafhaj, Jean Yim, Pascal Yim, Colin Noort. Prediction of HVAC System Parameters Using Deep Learning. 12th eSim Building Simulation Conference, Jun 2022, Ottawa (Ontario), Canada. ⟨hal-03930752⟩