Abstract
Several studies proved that occupant behaviour has a significant impact on energy consumption and indoor comfort. Thus, monitoring data and Internet of Things solutions are used to address behavioural changes for promoting energy efficiency and increasing, at the same time, the comfort perception. The possibility of seeing invisible information, such as energy consumption or comfort parameters, on a digital support have been proved to be effective in increasing users’ awareness and to encourage efficient behaviours. This paper presents a GUI developed for increasing the user awareness and involving them actively in addressing their actions and presents the back-end architecture for making prediction and sharing feedbacks with the users. By means of the interface user can: (1) Visualize information related to monitoring data, selecting, and filtering the data they would like to see. (2) Receive real time personalized feedbacks based on behavioural predictions defined by using Artificial Intelligence (AI) algorithm. The AI algorithms are based on a physics-assisted approach to achieve better results with less input. Missing (monitoring) data is calculated by applying physical models (building energy simulations) and only for the remaining parts machine learning models are used. Mainly we apply LSTM models. (3) Express personal comfort feedbacks based on comfort perception, for setting user-oriented feedbacks. The first part of the paper describes the architecture of the monitoring systems and presents the GUIs developed for two different case studies: a social housing building and a nursery school. The personalization of the GUIs based on user’s typology has been done for enhancing the active participation and the involvement of the users in the project. The second part of the paper presents the back-end architecture of the GUI and the AI algorithms used for monitoring data analysis. The physis-assisted algorithm allows us to make predictions based on occupancy behaviour and to provide each occupant with tailored personalized feedback to promote energy-saving behaviours in real-time. We have placed more than 150 sensors in these two buildings that return us almost 1000 measured variables that can be used for the training of the AI models.