Dataset Viewer
Auto-converted to Parquet Duplicate
ts
timestamp[us, tz=UTC]date
2026-02-05 23:57:25
2026-11-18 23:42:49
date
date32
state
stringclasses
51 values
county
stringlengths
3
53
county_fips
stringlengths
5
5
customers_out
int64
0
182k
customers_served
int64
1
1.27M
⌀
2026-02-05T23:57:25.756000
2026-02-05
AL
Autauga
01001
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Blount
01009
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Butler
01013
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Conecuh
01035
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Covington
01039
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Crenshaw
01041
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Dallas
01047
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
DeKalb
01049
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Etowah
01055
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Lowndes
01085
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Marshall
01095
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Monroe
01099
0
null
2026-02-05T23:57:25.756000
2026-02-05
AL
Wilcox
01131
0
null
2026-02-05T23:57:25.756000
2026-02-05
AK
Matanuska-Susitna
02170
0
8,039
2026-02-05T23:57:25.756000
2026-02-05
AK
Meadow Lakes/Houston
02170
0
5,200
2026-02-05T23:57:25.756000
2026-02-05
AK
Nancy Lake/Willow/Caswell Lks
02170
0
2,249
2026-02-05T23:57:25.756000
2026-02-05
AR
Arkansas
05001
0
10,160
2026-02-05T23:57:25.756000
2026-02-05
AR
Ashley
05003
0
8,552
2026-02-05T23:57:25.756000
2026-02-05
AR
Baxter
05005
0
7,817
2026-02-05T23:57:25.756000
2026-02-05
AR
Boone
05009
0
21,790
2026-02-05T23:57:25.756000
2026-02-05
AR
Bradley
05011
0
4,248
2026-02-05T23:57:25.756000
2026-02-05
AR
Calhoun
05013
0
571
2026-02-05T23:57:25.756000
2026-02-05
AR
Carroll
05015
0
4,185
2026-02-05T23:57:25.756000
2026-02-05
AR
Chicot
05017
0
5,292
2026-02-05T23:57:25.756000
2026-02-05
AR
Clark
05019
0
7,686
2026-02-05T23:57:25.756000
2026-02-05
AR
Clay
05021
0
3,119
2026-02-05T23:57:25.756000
2026-02-05
AR
Cleburne
05023
0
4,497
2026-02-05T23:57:25.756000
2026-02-05
AR
Cleveland
05025
0
1,372
2026-02-05T23:57:25.756000
2026-02-05
AR
Columbia
05027
1
12,638
2026-02-05T23:57:25.756000
2026-02-05
AR
Conway
05029
0
9,711
2026-02-05T23:57:25.756000
2026-02-05
AR
Craighead
05031
0
5,689
2026-02-05T23:57:25.756000
2026-02-05
AR
Crittenden
05035
0
12,084
2026-02-05T23:57:25.756000
2026-02-05
AR
Cross
05037
1
6,196
2026-02-05T23:57:25.756000
2026-02-05
AR
Dallas
05039
0
3,591
2026-02-05T23:57:25.756000
2026-02-05
AR
Desha
05041
0
5,857
2026-02-05T23:57:25.756000
2026-02-05
AR
Drew
05043
0
7,567
2026-02-05T23:57:25.756000
2026-02-05
AR
Faulkner
05045
0
29,383
2026-02-05T23:57:25.756000
2026-02-05
AR
Fulton
05049
0
852
2026-02-05T23:57:25.756000
2026-02-05
AR
Garland
05051
0
62,758
2026-02-05T23:57:25.756000
2026-02-05
AR
Grant
05053
0
5,948
2026-02-05T23:57:25.756000
2026-02-05
AR
Greene
05055
0
773
2026-02-05T23:57:25.756000
2026-02-05
AR
Hempstead
05057
0
799
2026-02-05T23:57:25.756000
2026-02-05
AR
Hot Spring
05059
0
16,592
2026-02-05T23:57:25.756000
2026-02-05
AR
Independence
05063
0
20,492
2026-02-05T23:57:25.756000
2026-02-05
AR
Izard
05065
0
3,785
2026-02-05T23:57:25.756000
2026-02-05
AR
Jackson
05067
0
5,017
2026-02-05T23:57:25.756000
2026-02-05
AR
Jefferson
05069
0
31,702
2026-02-05T23:57:25.756000
2026-02-05
AR
Lafayette
05073
0
2,368
2026-02-05T23:57:25.756000
2026-02-05
AR
Lawrence
05075
0
5,659
2026-02-05T23:57:25.756000
2026-02-05
AR
Lee
05077
0
3,169
2026-02-05T23:57:25.756000
2026-02-05
AR
Lincoln
05079
0
788
2026-02-05T23:57:25.756000
2026-02-05
AR
Logan
05083
0
4
2026-02-05T23:57:25.756000
2026-02-05
AR
Lonoke
05085
5
15,363
2026-02-05T23:57:25.756000
2026-02-05
AR
Marion
05089
0
10,834
2026-02-05T23:57:25.756000
2026-02-05
AR
Miller
05091
0
127
2026-02-05T23:57:25.756000
2026-02-05
AR
Mississippi
05093
0
14,553
2026-02-05T23:57:25.756000
2026-02-05
AR
Monroe
05095
0
3,905
2026-02-05T23:57:25.756000
2026-02-05
AR
Montgomery
05097
0
7,490
2026-02-05T23:57:25.756000
2026-02-05
AR
Nevada
05099
0
2,365
2026-02-05T23:57:25.756000
2026-02-05
AR
Newton
05101
0
529
2026-02-05T23:57:25.756000
2026-02-05
AR
Ouachita
05103
0
8,982
2026-02-05T23:57:25.756000
2026-02-05
AR
Perry
05105
0
105
2026-02-05T23:57:25.756000
2026-02-05
AR
Phillips
05107
1
8,793
2026-02-05T23:57:25.756000
2026-02-05
AR
Pike
05109
0
2,531
2026-02-05T23:57:25.756000
2026-02-05
AR
Poinsett
05111
0
8,607
2026-02-05T23:57:25.756000
2026-02-05
AR
Pope
05115
7
30,790
2026-02-05T23:57:25.756000
2026-02-05
AR
Prairie
05117
1
5,791
2026-02-05T23:57:25.756000
2026-02-05
AR
Pulaski
05119
9
158,505
2026-02-05T23:57:25.756000
2026-02-05
AR
Randolph
05121
0
3,217
2026-02-05T23:57:25.756000
2026-02-05
AR
St. Francis
05123
0
7,838
2026-02-05T23:57:25.756000
2026-02-05
AR
Saline
05125
0
20,603
2026-02-05T23:57:25.756000
2026-02-05
AR
Scott
05127
0
297
2026-02-05T23:57:25.756000
2026-02-05
AR
Searcy
05129
6
2,425
2026-02-05T23:57:25.756000
2026-02-05
AR
Sharp
05135
0
7,834
2026-02-05T23:57:25.756000
2026-02-05
AR
Stone
05137
0
8,560
2026-02-05T23:57:25.756000
2026-02-05
AR
Union
05139
0
23,613
2026-02-05T23:57:25.756000
2026-02-05
AR
Van Buren
05141
0
5
2026-02-05T23:57:25.756000
2026-02-05
AR
White
05145
5
34,685
2026-02-05T23:57:25.756000
2026-02-05
AR
Woodruff
05147
0
2,089
2026-02-05T23:57:25.756000
2026-02-05
AR
Yell
05149
0
11,394
2026-02-05T23:57:25.756000
2026-02-05
CA
Dehesa/ Alpine/ Rancho Palo Verde
06003
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Kern
06029
44
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Blossom Valley/ El Monte
06037
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Los Angeles
06037
4,294
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Mono
06051
8
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Orange
06059
272
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Riverside
06065
448
null
2026-02-05T23:57:25.756000
2026-02-05
CA
San Bernardino
06071
564
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Clairemont/ Kearny Mesa
06073
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Dulzura/ Potrero/ Campo
06073
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Granite Hills/ Bostonia/ East El Cajon
06073
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Julian/ Santa Ysabel/ Morettis
06073
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
La Jolla/ La Jolla Shores/ Soledad Mountain
06073
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Live Oak Springs/ Boulevard/ Jacumba
06073
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
North City West/ Torrey Pines/ Fairbanks Country Club
06073
26
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Ocean Beach/ Midway/ Lindbergh Field
06073
70
null
2026-02-05T23:57:25.756000
2026-02-05
CA
San Diego
06073
16
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Sorrento/ Mira Mesa/ Scripps Ranch
06073
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Torrey Pines/ Sorrento Valley/ University City
06073
0
null
2026-02-05T23:57:25.756000
2026-02-05
CA
Santa Barbara
06083
8
null
End of preview. Expand in Data Studio

US power outages by county, every 20 minutes

Customers without power in US counties and states, as reported on electric utilities' public outage maps, collected every 20 minutes since 2026-02-05 by The Observatory. One Parquet file per UTC day, added nightly.

  • county (default): one row per county a utility reported in a scrape run, with its 5-digit FIPS code. About 110,000 rows a day.
  • state: one row per state per scrape run (50 states, DC and the US Virgin Islands). About 3,700 rows a day.
from datasets import load_dataset

county = load_dataset("stevensonl/observatory-dataset", "county", split="train")
state = load_dataset("stevensonl/observatory-dataset", "state", split="train")
import duckdb  # one day, straight from the Hub

duckdb.sql("""
  SELECT county_fips, max(customers_out) AS peak_out
    FROM 'hf://datasets/stevensonl/observatory-dataset/data/county/2026/2026-11-17.parquet'
   GROUP BY 1 ORDER BY 2 DESC LIMIT 10
""")

Files

data/county/<year>/<YYYY-MM-DD>.parquet
data/state/<year>/<YYYY-MM-DD>.parquet

Each file holds the scrape runs whose timestamp falls in that UTC day (normally 72). A day's file is written once, after the day ends, and is not revised. A day with no file is a day the scraper did not run (see Known gaps).

Columns

county

Column Type Meaning
ts timestamp (UTC) When the scrape run started. Every row of one run has the same ts, so it identifies the run. Runs start at :00, :20 and :40 and are stamped a few minutes later.
date date The UTC day of ts (the file's day).
state string Two-letter state code.
county string County name as resolved from the utility's label.
county_fips string 5-digit county FIPS code (2-digit state + 3-digit county), 2022 vintage. Always present; leading zeros kept, so read it as text.
customers_out int64 Customers without power in the county at that run, summed over the utilities that reported it. A customer is a meter or account, not a person. 0 means a utility reported the county with nobody out.
customers_served int64, nullable Customers those utilities say they serve in the county. Null when their maps give no count. It is a reporting denominator, not the county's total customer base.

(ts, state, county) is unique. (ts, county_fips) is unique in almost every run; on rare occasions two labels resolve to the same county in one run and both rows are kept (about 0.02% of rows in October 2026).

state

Column Type Meaning
ts, date, state As above.
customers_out int64 Customers without power in the state at that run, from the utilities' own totals.
customers_served int64, nullable Customers served that the utilities reported for the state.
customers_out_no_county int64 Customers out that utilities reported with no county the scraper could resolve. These are missing from the county subset; this column is how much.

State totals are built from utility-level totals, not by adding up county rows, so a state's customers_out can differ a little from the sum of its counties in the same run.

How to read it

  • A missing county row is not a zero. A county has a row in a run only if a tracked utility reported it. No row means "not reported", which is unknown, not "no outages".
  • Coverage is partial and uneven. A county's figure covers only the utilities that are tracked and that name the county. Where a co-op or municipal utility is not tracked, its customers are missing from both customers_out and customers_served.
  • These are utility-reported snapshots, not measurements and not event records. Utilities update their maps on their own schedules, round or suppress small numbers differently, and sometimes publish stale or wrong figures. Nothing here is corrected or smoothed.
  • No outage causes, durations or locations below county level are included, and no cost estimates.

Known gaps

  • Counties that never report. As of 2026-11-16, 2,483 of 3,144 counties had reported at least once; 661 never had. About half are served only by utilities that are not tracked yet (mostly rural co-ops and public power districts in Nebraska, Kansas, Iowa, Montana and the Dakotas). The rest are served by a tracked utility whose map gives outage points or service areas instead of counties. That includes large ones: PG&E territory in California, Consumers Energy in Michigan, APS in Arizona and Con Edison in New York City mostly land in customers_out_no_county, and Maricopa County, AZ has never had a county row.
  • Connecticut. The 2022 FIPS codes use nine planning regions; the utility reports the eight former counties, so most of Connecticut has no county rows.
  • Utilities that were failing. Some utilities' feeds were unreachable for stretches, including the Duke Energy companies (NC, SC, FL, IN, OH, KY) and several Avangrid companies (ME, NY) in autumn 2026. Their counties are absent or undercounted in those periods; no row marks the difference.
  • Days with no data: 2026-04-05 to 2026-04-10, 2026-09-01 and 2026-09-07. Other days can have fewer than 72 runs; count distinct ts to check.
  • Territories. Puerto Rico, Guam and the other territories are not covered. The US Virgin Islands appear in state only.
  • The set of utilities grows. Utilities were added and fixed over time, so an upward step in a county's or state's series can be new coverage rather than a new outage. Compare customers_served across time before reading a trend.

Sources and attribution

The figures are county and state aggregates derived from the public outage maps of about 260 US electric utilities, read with ordinary web requests every 20 minutes. Utility-level and outage-level records are not redistributed here. The underlying facts belong to the utilities that publish them; this dataset is not affiliated with or endorsed by any of them. County names and FIPS codes follow the US Census Bureau's 2022 county definitions.

Collected and published by Luke Stevenson as part of The Observatory.

Licence

CC BY 4.0. Please credit "The Observatory (observatory.luketstevenson.com)" and link to this dataset. The data is provided as is, with no warranty of accuracy or completeness; do not use it for safety-critical decisions. If you operate a utility and want your figures handled differently, open a discussion on this repository.

Citation

@misc{observatory_power_outages,
  author = {Stevenson, Luke},
  title  = {US Power Outages by County, every 20 minutes},
  year   = {2026},
  url    = {https://huggingface.co/datasets/stevensonl/observatory-dataset}
}
Downloads last month
-