Overview of the challenge: "Restoring power after an outage event is a critical piece of response and recovery for those in the emergency response sector. When analyzing data on power outages, we seek to understand spatial and temporal patterns, restoration times, correlations with other relevant datasets, future outage predictions, expected outage behavior, and how we might fill in data gaps. Analysis at the county level is recommended when possible."
More details about the challenge can be viewed here: https://smc-datachallenge.ornl.gov/eagle/
The Environment for the Analysis of Geo Located Energy Information (EAGLE-I) monitors the number of customers experiencing outages at the county level across the United States. Covering 92% of US and territories, this dataset provided by ORNL provides a comprehensive, real-time source of outage information. The following information is provided:
eaglei_outages_2014.csv,eaglei_outages_2014.csv,eaglei_outages_2015.csv,eaglei_outages_2016.csv,eaglei_outages_2017.csv,eaglei_outages_2018.csv,eaglei_outages_2019.csv,eaglei_outages_2020.csv,eaglei_otuages_2021.csv,eaglei_outages_2022.csv-
Provides outage data for each county at intervals of 15 minutes. Entries where the value is 0 are omitted.
-
Columns:
Name Pandas data type Description fips_code int64 County FIPS code county object County name state object State name sum float64 The number of customers out of power. Can be converted to int64. run_start_time object Timestamp at which outages were recorded. Can be converted using pd.to_datetime()
-
coverage_history.csv- Provides information about the minimum and maximum number of customers covered for each state for each year included in this dataset.
- Columns:
Name Data type Description year object Date the information was recorded. Can be converted using pd.to_datetime()state object The two character code for the state total_customers int64 The number of customers in the state for the year min_covered int64 The minimum number of customers covered max_covered int64 The maximum number of customers covered min_pct_covered float64 The minimum percent of customers covered max_pct_covered float64 The maximum percent of customers covered
- Calculate outage durations for each county
- Find correlations between weather and social vulnerability datasets to corresponding outage data
- Forecast future outages
To download the datasets from the following links, you will need to use Globus and create an account.
-
If you are using an HPC: Once you sign in, check if there is a Globus collection hosted on your HPC in the 'File Manager' tab.
-
In most other cases, install Globus Connect Personal on whichever device you would like to have these datasets downloaded to. Once your server is running, you can search for it like any other connection.
After Globus is set up, these links will help you search for download the dataset via Globus:
- EAGLE-I Umbrella DOI: https://doi.ccs.ornl.gov/ui/doi/436
- 2014-2022 DOI: https://doi.ccs.ornl.gov/ui/doi/435
- US Weather Events (Kaggle)
- National Weather Service Valid Time Extent Code (VTEC) Archives
- Climate and Economic Justice Screening Tool
These can be installed using pip, conda, or mamba.
- Dataframes:
vaex,pandas,geopandas - Math:
numpy - Time series modeling:
darts,statsmodels - Forecasting:
prophet,xgboost,sklearn,pytorch - Plotting:
matplotlib,seaborn