|
 |
|
|
|
A Machine Learning Approach to Microclimate Monitoring and Fault Detection |
|
PP: 327-334 |
|
doi:10.18576/amis/190209
|
|
Author(s) |
|
Nurkamilya Daurenbayeva,
Lyazzat Atymtayeva,
Almas Nurlanuly,
Artem Bykov,
Bakhytzhan Akhmetov,
Gabit Shuitenov,
Umut Turusbekova,
|
|
Abstract |
|
The integration of Machine Learning (ML) into heating, ventilation, and air conditioning (HVAC) systems significantly enhances fault detection and diagnosis (FDD), crucial for improving energy efficiency, as buildings account for around 40% of global energy consumption. However, the presence of inaccurate and noisy data can complicate FDD efforts, as traditional methods often struggle in real-world conditions. This study explores FDD through an experiment conducted in two distinct environments: residential and non-residential buildings. Using a Node MCU microcontroller and over 16 sensors, data on microclimate parameters such as temperature, humidity, and CO2 levels were collected and analyzed in real time. The findings highlighted the variability of microclimate conditions and identified challenges associated with existing FDD methods, including the limitations of Principal Component Analysis (PCA) in noisy environments. Recent literature categorizes ML-based FDD methods into three groups: traditional machine learning, deep learning, and hybrid models, demonstrating their superiority over conventional approaches. However, challenges such as data variability and the need for real-time processing still exist. To develop intelligent fault diagnosis systems, the CRISP-DM methodology is proposed, encompassing phases from business understanding to deployment, while addressing potential noise and inaccuracies. The system architecture includes sensors for monitoring climate and air quality, a microcontroller for data processing, a user interface for real-time notifications, and analysis algorithms for anomaly detection. Overall, this research underscores the potential of ML in optimizing Heating, Ventilation, and Air Conditioning performance and emphasizes the need for adaptive models and IoT integration to enhance data collection efficiency, marking an important step toward sustainable energy practices in building microclimate control.
|
|
|
 |
|
|