An Indoor Air Temperature Prediction Framework for Model Predictive Control in HVAC Systems

Abstract

An accurate prediction of indoor air temperature (IAT) for heating, ventilation, and air conditioning (HVAC) systems in smart buildings is a challenging task, especially for a large future prediction horizon. In particular, the dependency of IAT modelling on various complex factors including outdoor ambient conditions, a large set of HVAC system parameters, and the nonlinearity which exists in building’s thermodynamics make an accurate prediction of IAT difficult. Herein, we present a black-box model based IAT prediction algorithm which is used for an accurate IAT prediction using an artificial intelligence and a data-driven approach. Towards this, a Bi-LSTM (Long-short term memory) based model is proposed which is also compared with a standard LSTM based model. Our experimental results showed an improvement in IAT prediction of up to 10% using our proposed model when compared with a standard LSTM model for IAT prediction. These results are significant for applications where accurate IAT predictions can be used for a more robust energy consumption modelling and optimization in large scale buildings.