Green IoT Event Detection for Carbon-Emission Monitoring in Sensor Networks

Area of Science:

• Environmental Science
• Computer Engineering
• Sustainable Technology

Background:

• Traditional AI/ML platforms for event detection are resource-intensive, leading to high power consumption and carbon emissions.
• Limited resources in embedded systems pose challenges for real-time data processing and event classification.
• There is a need for energy-efficient solutions for monitoring environmental factors like carbon emissions.

Purpose of the Study:

• To introduce an innovative approach using low-power microcontrollers for real-time binary classification of events.
• To demonstrate the efficiency of microcontrollers in processing sensor data with limited resources.
• To showcase significant power savings and carbon emission reduction compared to traditional AI/ML platforms.

Main Methods:

• Utilizing homogeneous hardware/firmware for real-time event detection on microcontrollers.
• Implementing binary classification algorithms suitable for resource-constrained environments.
• Conducting case studies on landfill CO2 emissions and home energy usage monitoring.

Main Results:

• Achieved significant power savings (94.8-99.8%) by minimizing data transmission during non-event periods.
• Demonstrated the feasibility and effectiveness of microcontroller-based event detection.
• Highlighted the substantial reduction in power draw and carbon emissions compared to conventional AI/ML systems.

Conclusions:

• Low-power microcontrollers offer an efficient and sustainable solution for real-time event detection and classification.
• This approach significantly reduces energy consumption and environmental impact.
• Microcontroller-based systems provide a viable alternative to resource-intensive AI/ML platforms for carbon emission reduction efforts.