Project Description

Author: Benedikt Herold, benedikt.herold@tuwien.ac.at

License: Creative Commons Attribution 4.0 International

DOI: 10.70124/xzk92-8mm08

DMP Reference: 10.5281/zenodo.17732654

Description on the Dataset

This is an detailed description on the data provided via the csv files, which can be read with any csv-interpreter software.

Time is given in the standard python datetime format, all other values except ToI refer to power values with the unit W. ToI is unit less and a logical counter value.

Description for "Filter_Difference_Calculation.csv"

This file contains the raw and processed power data for the aggregated profile and individual devices. The columns include time in Python datetime format, Raw_Power representing the unprocessed aggregated power data, Filtered_Power which is the median-filtered signal (window size 3) to reduce spikes while preserving sharp edges, and First_Order_Difference which is the first-order difference of the filtered signal to extract step sizes. Additionally, it contains the filtered measurements for individual devices Measured_Device_Power, and ToI is a boolean mask marking Timestamps of Interest (ToIs) that were detected.

Description for "Event_Detection_Matching_Thresholds.csv"

This file captures the generated event maps based on different matching score thresholds. The columns include time in Python datetime format, High_Threshold , Optimal_Threshold, and Low_Threshold, representing the event maps generated at corresponding confidence levels, and ToI, a boolean mask indicating Timestamps of Interest (ToIs) detected in the event maps.

Description for "Reconstructed_Detected_Grid_Power.csv"

This file provides reconstructed and filtered measurements for both aggregated and individual devices. The columns include time in Python datetime format, Filtered_Aggregated_Power representing the filtered aggregated measurements, Reconstructed_Power which is the reconstructed power profile of detected devices, Measured_Power for the filtered measurements of individual devices, Remaining_Unlabeled_Power indicating remaining unlabeled power, and ToI, a boolean mask marking Timestamps of Interest (ToIs) that were detected.

Instruction for running the project code

For generation of the described dataset execute the provided python file edge_detection.py without any additional arguments.

In order du run the script following libraries are required:

  • pandas
  • numpy
  • matplotlib
  • scipy