HIFDA - High-Frequency Electrical Voltage and Current Signals from Household Appliances.

Area of Science:

• Electrical Engineering
• Computer Science
• Energy Systems

Background:

• Smart grids require efficient energy management through Non-Intrusive Load Monitoring (NILM).
• Existing NILM datasets have limitations like low sampling rates and lack of steady-state data, hindering appliance detection.
• High sampling rates are crucial for accurate appliance classification and identifying low-power devices.

Purpose of the Study:

• Introduce the HIgh Frequency household electrical signals DAtaset (HIFDA) for advanced NILM research.
• Provide a high-frequency dataset suitable for training and validating machine learning models in NILM applications.
• Address the limitations of current datasets by offering high sampling rates and steady-state data.

Main Methods:

• Collected high-frequency (100 kSPS) steady-state electrical signals from 14 household appliances and an empty grid.
• Utilized a custom System-on-Chip (SoC) device for data acquisition.
• Focused on active power consumption and incorporated multiple time windows for comprehensive analysis.

Main Results:

• Developed HIFDA, a novel dataset with high-frequency electrical signals.
• The dataset captures steady-state consumption patterns essential for NILM.
• HIFDA includes data from 14 appliances and the empty grid, offering diverse usage scenarios.

Conclusions:

• HIFDA is a valuable resource for advancing NILM techniques and smart grid energy management.
• The high sampling rate and steady-state data in HIFDA improve appliance classification accuracy.
• This dataset supports research in NILM and other applications requiring detailed electrical signal analysis.