Non-intrusive occupancy analysis in the Musée L

Details

Supervisors

Pelsser, Cristel ; Sadre, Ramin

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en sciences informatiques, à finalité spécialisée

Abstract

enOccupancy detection is a field of study that gains particular interest in the context of smart buildings with various applications, such as reduction of Heating, Ventilation, and Air Conditioning (HVAC) energy consumption, security, and comfort of the occupant indoors. This thesis focuses on non-intrusive occupancy detection in the Musée L, a museum located in Louvain-la-Neuve. Therefore, it prioritizes the visitors’ privacy by not using any cameras and aims to be low-cost by utilizing already installed sensors. The objective is to analyse occupancy estimation models through environmental sensors and finding the most accurate machine learning model. A system was developed to measure the Indoor Air Quality (IAQ) using Internet of Things (IoT) devices set up with multiple sensors for carbon dioxide (CO2), temperature and humidity. Data was transmitted via the MQTT protocol to the central MQTT broker. For every experiment, the occupancy ground truth was observed and used as output in the machine learning models chosen for training to predict the occupancy range. With the best model, the bagging classifier, we achieved an accuracy of 77.5% and employed this model to obtain the closest estimation of occupancy in the museum with the historical data collected by the IAQ sensors of the GTPL from 2022 to 2024. To complete this project, we used tools such as ESP-IDF, the framework used to program the IoT devices, ESP32s3, in C. For the machine learning model, we coded in Python with the libraries Scikit-learn and Pandas to process the data.