All demo datasets

Electric — Municipal Utility

Greenfield Municipal Power (synthetic)
Guided tour Sample data · no sign-up
Group on Customer class Geography Service type Month Quarter Year
UnitTrack · Disclosure-controlled report

Electric — Municipal Utility

Greenfield Municipal Power (synthetic)
Generated 2026-07-27 19:36 · Grouped by customer class · k = 15
Residual re-identification risk
worst case 3% · Low · smallest published group 37
3/3
Customer class groups published
100.0%
of meter reads in published output
0
groups withheld by disclosure control
15/15 rule
≥15 contributors; no single customer over 15%
Contents
  1. Disclosure control & data provenance
  2. Consumption profile
  3. Geography & spatial clustering
  4. Weather & seasonality
  5. Demand & load
  6. Benchmarking (EUI)
  7. Energy burden & equity
  8. Measurement & verification (M&V)
  9. Differential privacy (alternative release)
  10. Appendix — policy & citations

Disclosure control & data provenance

Grouping the dataset into 3 customer class groups, 3 were safe to publish and 0 were withheld under the rules below. Across 18000 meter reads, 100.0% of records and 100.0% of total total consumption remain in the published output, with worst-case re-identification risk of 3% (Low).

Data quality & reliability
18000
Loaded reads
0 rejected & quarantined
4.0%
Estimated reads
utility estimate flag
0.3%
Negative usage
rollovers / net metering
2763
High outliers
Tukey 1.5·IQR (1977)

Anomalies are flagged, not dropped; rows failing validation are quarantined as rejected. 156 reads (0.9%) lack coordinates and are excluded from spatial views.

Methods & authorities
  • No groups required suppression at these settings (k = 15).
  • PII handling — identifiers are excluded from the analytic path; only disclosure-controlled aggregates are reported.
    NIST SP 800-122; NISTIR 8053
Published results
Customer class Total consumption (kWh) Contributors Status
INDUSTRIAL 54,388,597 37 Published
COMMERCIAL 39,401,813 233 Published
RESIDENTIAL 27,296,524 1230 Published
Explore live →

Consumption profile

Total consumption by customer class — published groups (kWh)
Consumption distribution & inequality
0.739
Gini coefficient
Gini 1912; Lorenz 1905
3.708
Coefficient of variation
std ÷ mean
21721
Median use / account (kWh/yr)
across 1500 accounts
230
High outliers
Tukey 1.5·IQR
Lorenz curve — cumulative use vs. cumulative accounts
Percentiles (kWh/yr)
10th10315
25th14478
Median21721
75th39206
90th140491
Customer class × month
Total consumption — class (rows) × month (columns); blanks were suppressed (kWh)
Explore distribution live →

Geography & spatial clustering

Random
Spatial pattern
not significant
0.147
Moran's I
expected -0.125
0.65
z-score
over 9 areas
0.5168
p-value
two-tailed
Total consumption by area (kWh) — uncolored areas are suppressed

No significant spatial pattern — values look randomly placed.

1 hot spot (Getis-Ord Gi*, p < 0.05).

Global Moran's I, rook contiguity; normality-assumption inference. Moran (1950); Cliff & Ord (1973). Basemap © OpenStreetMap/Carto.
Explore spatial live →

Weather & seasonality

7.3%
Weather-sensitive (cooling-driven)
ASHRAE G14; PRISM
0.0% / 100.0%
Heating / cooling split
of weather-driven use
—
Weather-normalized annual
restated at normal year
0.952
Model fit (R²)
degree-day regression

E = 7039 + -4.2528·HDD + 2.4549·CDD  ·  base 65°F · 11 months

Seasonal use — 37.6% above winter baseline, peak Aug
Average monthly consumption vs. winter baseline (kWh)
Winter-baseline (minimum-month) method. Standard seasonal decomposition.
Explore weather live →

Demand & load

0.284
Median load factor
across 1500 buildings
4921
Coincident peak (kW)
hour 10:00
0.133
Coincidence factor
NREL UMP Ch.10
7.53
Diversity factor
1 ÷ coincidence
Peak demand by hour of day (summed customer peaks; tallest = system coincident peak)
Load factor by class
ClassBuildingsp25Medianp75
RESIDENTIAL 1230 0.2330.2710.307
COMMERCIAL 233 0.3680.4050.447
INDUSTRIAL 37 0.480.5160.579
Load factor: utility load research; NREL UMP. Coincidence/diversity: NREL UMP Ch.10; BPA M&V. Small classes withheld (k-anonymity).
Explore demand live →

Benchmarking — Energy Use Intensity

35.7
Median EUI (kBtu/ft²/yr)
EPA ENERGY STAR; DOE/EIA CBECS
1500
Buildings benchmarked
floor area & ≥ 2 months
1 kWh = 3.412 kBtu; 1 therm = 100 kBtu (EIA)
reference constants v2024.1
EUI distribution across buildings (published classes)
By customer class (site EUI, kBtu/ft²/yr)
ClassBldgsp25Medianp75vs nat'l
RESIDENTIAL 1230 23.436.457.4 121%
COMMERCIAL 233 15.529.257.4 42%
INDUSTRIAL 37 19.253.7131.2 —
Within-dataset EUI percentiles vs. class peers. “vs nat'l” is coarse all-fuel context (CBECS 2018 / RECS 2020), not the ENERGY STAR 1–100 score. Small classes withheld.
Explore benchmarking live →

Energy burden & equity

Burden = annual energy cost ÷ tract median household income at $0.16/kWh. Bands: ≤6% affordable · 6–10% moderate · >10% high (ACEEE).

4.69%
Median tract burden
across 46 tracts
3
High-burden tracts
> 10% of income
23.9%
Households high-burden
300 of 1253
31
Tracts suppressed
below min-count
Energy burden by census tract (% of income)
Energy burden = annual energy cost ÷ median household income; ≤6% / 6–10% / >10% cut points (6% affordability from the Home Energy Affordability Gap [Fisher, Sheehan & Colton]; used by DOE/NREL LEAD & ACEEE — convention, not statute). Income: ACS 5-year B19013. Rate: EIA average retail price. Tract-level per DOE/NREL LEAD; metered-commodity burden is a lower bound on total home-energy burden.
Explore burden live →

Measurement & verification (IPMVP Option C)

Baseline model fit on 5 months before Jul 2023; savings measured over the following 6 reporting months. Synthetic demo with no intervention — avoided energy near 0% is the method working correctly.

-1.3%
Avoided energy
-549 kWh/meter
41891
Adjusted baseline (CBL)
kWh/meter expected
42440
Reporting actual
kWh/meter metered
0.953
Baseline fit (R²)
degree-day model
Reporting period: metered actual vs. adjusted baseline (kWh/meter)
IPMVP Option C (EVO 10000-1:2016); ASHRAE Guideline 14; base-65°F degree days (US EIA / NOAA).
Explore M&V live →

Differential privacy — an alternative release

An alternative to suppression: instead of hiding small groups, calibrated Laplace noise is added to each count (ε = 1.0, scale b = 1.0) so no individual's presence can be inferred while totals stay useful. Mean absolute error ±0.7 across 3 groups.

True vs. differentially private counts by customer class
Laplace mechanism, ε-differential privacy. Dwork, McSherry, Nissim & Smith (2006); Dwork & Roth (2014); US Census Bureau 2020 DAS.
Explore differential privacy live →

Appendix — policy & citations

Disclosure-control settings
Governing standard15/15 rule
Minimum group size (k)15
Dominance (top-n ≤ p)1 ≤ 0.15
Min accounts for models15
GroupingCustomer class
Standards & sources cited
  • k-anonymity — Sweeney (2002); ESSnet/Eurostat SDC Handbook
  • (n,k)-dominance — ESSnet SDC Handbook §4.2; CPUC 15/15
  • Inequality — Gini (1912); Lorenz (1905)
  • Spatial — Moran (1950); Cliff & Ord (1973)
  • Weather / M&V — ASHRAE Guideline 14; IPMVP Option C; PRISM
  • Benchmarking — EPA ENERGY STAR; DOE/EIA CBECS/RECS
  • Load — NREL UMP; BPA M&V Peak Demand
  • Burden — ACEEE; US Census ACS
  • Differential privacy — Dwork et al. (2006); US Census 2020 DAS
  • PII — NIST SP 800-122; NISTIR 8053; NIST SP 800-188
  • Methods & Formulas — every rule & analysis with its formula-as-implemented and primary citation

Synthetic sample data — for demonstration, not distribution. Generated by UnitTrack, a Polish Clover Solutions product · 2026-07-27 19:36.