CoryVu — Overnight versus midday carbon intensity
Evidence edition: pilot1-method2-20260924
Observations: 2018-01-01 through 2025-12-31, GB.
Source: NESO National Carbon Intensity Forecast
https://www.neso.energy/data-portal/national-carbon-intensity-forecast
Source licence: https://www.neso.energy/data-portal/neso-open-licence
Supported by National Energy SO Open Data
NESO has not commissioned, endorsed or verified CoryVu's analysis.

CoryVu derived analysis, not raw source data. date_utc is the UTC calendar date.
Overnight interval starts: 00:00, 00:30, ..., 05:30. Midday: 11:00, 11:30, ..., 16:30.
Each block contains 12 half-hour observations; both blocks are end-exclusive.
Mean actual intensity per block, in gCO2/kWh. Difference = overnight minus midday.
Positive difference means overnight higher. Both block means require all observations.
No forecast substitution, interpolation, or missing-as-zero treatment.
Outcome uses exact source precision; displayed decimals use ROUND_HALF_UP to 3 places.
The *_exact fields are reduced rational numerator/denominator values from verified results.
The columns ending gco2_per_kwh are rounded decimal renderings, not new calculations.
The outcome_at_source_precision column is LOWER, HIGHER or EQUAL.
The evidence_edition column identifies the verified dataset edition in each record.

exclusions.csv: one row per excluded date. selected_source_rows counts selected source
records. issues_json lists every affected timestamp and verified reason. Multiple issues
on a date do not create additional excluded dates. Blank/missing values are never zero.
Excluded dates belong to neither outcome. No raw research paths or raw source records
are included. All exclusions for this edition are documented in the file.

Limitations: estimated average generation intensity, not marginal household emissions.
Historical frequency is not future probability. Fixed UTC periods are one hour later
in UK local time during BST. No current/tonight recommendation, electricity-price
conclusion or personal carbon-saving claim. Equal is equal at source precision.
Missingness is concentrated in 2023. Source estimates and later revisions may differ.
The former undocumented 0 <= actual < 2000 guard affected zero observations in this
edition. Method 2.0 records it as an anomaly/HOLD trigger, not a scientific validity range.

Source retrieved: 2026-09-23T15:09:43.721408+00:00
Source last-modified header: Wed, 23 Sep 2026 14:11:28 GMT
Frozen source SHA-256: 867d8cc909e4d47fe61c26c79d4f9a36fff488ff1bbc36a1aa52018a6145b117
Written method: 2.0; SHA-256 865582facdb63e2b46dc3769ead6ca96ba09785463f5859bb1dd9dd25e4952c6
Verification run: 2026-09-24T09:22:39.830503+00:00
Automated verification attestation SHA-256: ff4a5dd7fedc6c979ef4be41be238e0d84d48ed716efded07d0289031245b67f
Derived result SHA-256: 26561418d993aa3164e159e700c01621aef37db5b495fcf587b2cf03499c16fe
No independent expert or human computational review is claimed.