Occupancy Level Estimation in an Indoor Environment for Building Automation

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

Sähkötekniikan korkeakoulu | Master's thesis

Department

Mcode

ELEC3025

Language

en

Pages

70

Series

Abstract

This thesis explores the development of predictive models for estimating occupancy levels based on environmental data. Unlike conventional presence estimation relying solely on passive infrared (PIR) sensors, this study investigates the feasibility of employing a combination of low-cost sensors, including carbon dioxide (CO2), volatile organic compounds (VOCs), sound, temperature, light, and humidity, alongside PIR sensors. The thesis employs a systematic approach encompassing hardware setup, data collection, data pre-processing, data labeling, data visualization, exploratory analysis, modeling using classical Machine Learning techniques, and evaluation. Real-world data collected from two meeting rooms of different sizes and capacities, Dartford (4-person capacity) and Tampere (9-person capacity), was utilized to validate the proposed methodology. Extensive analysis identified CO2, PIR, VOC, and sound levels as the most influential environmental parameters for occupancy estimation. Remarkable accuracies of 97.5% and 97.7% were achieved for presence estimation in the Dartford and Tampere meeting rooms, respectively. Similarly, occupancy level estimation yielded accuracies of 95.35% and 92.8% in the Dartford and Tampere meeting rooms, respectively. The findings demonstrate the potential of using environmental parameters for predicting occupancy levels, paving the way for building automation and enhanced energy savings based on occupancy information.

Description

Supervisor

Xiao, Yu

Thesis advisor

Takala, Pasi
Juslén, Henri

Other note

Citation