Multi-Sensor Platform for Predictive Air Quality Monitoring

Area of Science:

• Indoor air quality monitoring
• Environmental sensing and control
• Internet of Things (IoT) applications

Background:

• Carbon dioxide (CO2) significantly impacts indoor air quality and occupant health.
• Traditional air quality control systems require extensive data over long periods, limiting adaptability.
• Dynamic changes in occupant behavior and environmental conditions necessitate adaptive monitoring solutions.

Purpose of the Study:

• To develop an adaptive IoT platform for accurate, short-term CO2 concentration forecasting.
• To enable proactive control of HVAC systems for enhanced indoor air quality and energy efficiency.
• To overcome the limitations of traditional methods requiring long-term data training.

Main Methods:

• Development of an adaptive hardware-software platform based on the IoT paradigm.
• Utilizing a limited window of recent data for CO2 trend analysis.
• Evaluation of three deep-learning algorithms, including Long Short-Term Memory (LSTM) networks.
• Testing in a real-world residential setting with parameters including physical activity, temperature, humidity, and CO2 levels.

Main Results:

• The developed IoT system achieved high accuracy in forecasting CO2 trends.
• The Long Short-Term Memory (LSTM) network demonstrated superior performance among the evaluated algorithms.
• An approximate Root Mean Square Error (RMSE) of 10 ppm was achieved with a training period of only 10 days.

Conclusions:

• The adaptive IoT platform effectively forecasts indoor CO2 levels using minimal recent data.
• LSTM networks provide a robust solution for accurate and timely air quality prediction.
• The system offers a cost-effective and responsive approach to managing indoor air quality and HVAC systems.