PowerGridIQ · Methodology · Cost engine
Every figure we have published, withdrawn and replaced, with the reason and the replacement. We publish corrections rather than quietly changing the wording.
This is the only page on the site where a superseded figure still appears. Everywhere else — market pages, APIs, exports, structured data and machine-discovery documents — it has been removed, because a correction that restates the wrong number in a machine-readable field has republished it in the one place a scraper reads.
Corrected 2026-08-24 · defect #320 · ISO New England
| Withdrawn 2026-08-24 | Replacement |
|---|---|
triangulated Published from initiation until 2026-08-24 | $None per MWh partial_regional_only tariff_derived, regional layer only |
Not the arithmetic -- the scope. A single delivered-cost estimate was published for the whole ISO New England footprint, as though one customer existed there. The footprint contains 32 serving utilities across six states under five regulators, including roughly twenty municipal utilities not commission-regulated for retail. Two 100 MW loads, one in Boston on Eversource Massachusetts and one in Maine on Versant, share an identical regional charge and can differ materially below it. One number cannot be right for both, and the row had no way to say which it meant.
The regional layer, independently derived and published as a verified partial: $21.34/MWh Regional Network Service at the filed Pool RNS Rate of $177.62722/kW-yr, plus $0.23/MWh OATT Schedule 1 at $1.87420/kW-yr, giving $21.57/MWh at the standard 100 MW, 95% load-factor archetype. This is NOT a delivered cost and is never published as one.
The cost pillar moves to not_assessed. The verified $21.57 is deliberately NOT fed into it: a partial figure entering a delivered-cost rubric would score as though regional transmission were the whole bill, which would raise the pillar for a market that just lost its cost estimate. The published methodology defines no reweighting for an unavailable pillar, so no replacement score is manufactured. The overall rating is placed under review and the prior score and tier are preserved in history as superseded.
Defects closed:
Evidence: ISO-NE General Business Practices Section 2, version 31 July 2026; OATT Schedule 9 settlement description and calculation summary v4.0; ISO-NE Transmission Service Types.
Source
Corrected 2026-08-23 · defect #319 · Northern Virginia (PJM)
| Withdrawn 2026-08-23 | Replacement |
|---|---|
triangulated Published from initiation until 2026-08-23 | $82.63 per MWh tariff_point administered |
The band was built by taking a Dominion-zone wholesale energy price and adding capacity, network and ancillary estimates on top. That construction suits a market where a large customer buys energy competitively. It does not suit Virginia, where Schedule GS-4 is a bundled cost-of-service tariff and the customer never pays a wholesale zone price.
The record declared a spot anchor of $38/MWh and an energy line of $96/MWh, and its own note said the energy line WAS the Dominion-zone wholesale. A line asserted to be the wholesale price cannot be 2.53 times the wholesale price recorded in the same object. Nothing outside the record was needed to see it.
A build from Schedule GS-4 at transmission service voltage, including all fifteen riders in the current Exhibit of Applicable Riders and applicable Virginia taxes, gives $82.63 per MWh for the standard case of 100 MW at a 95% annual load factor.
A qualifying data centre also pays Virginia's data-centre electricity tax of $0.011/kWh, adding about $11.00 per MWh and giving $93.63 per MWh before any refund. That is a DIFFERENT customer from the published archetype, which is a general large load and does not automatically incur the tax.
Rider OSW is worth about $2.82 per MWh, and its exemption requires a qualifying 15-year offshore-wind procurement arrangement. It is published as an ADJUSTMENT only, never as a delivered-cost total: removing the rider without adding the separately contracted cost of offshore-wind supply and environmental attributes would understate what that customer actually pays. The net delivered cost for such a customer is not publicly observable.
Published from initiation until 23 August 2026, on the market page, in the ratings and screening APIs, in shortlists, market memos and structured exports. Anyone who screened for markets under a cost ceiling between roughly $83 and $130 per MWh may have seen Northern Virginia excluded when a correctly built figure would have included it.
The cost pillar moves from 70 to 74 and the overall score from 51 to 52. The tier is unchanged at Tier 2 Strong. The move is small because the cost pillar was never a function of the band: under the now-declared cost-pillar rubric the withdrawn $130 band scores 52, against a pillar that was actually published at 70.
Every other banded market was searched for the same construction, using a mechanical test: an energy line materially above that market's own wholesale anchor, in a record claiming the line is wholesale. Two markets showed it, Northern Virginia and the Pacific Northwest, and both were already withdrawn. Two others diverge from their anchor and explain why in their own records, which is a disclosed modelling choice rather than a contradiction. No further figures are affected and none was changed silently.
Defects closed: #307, #308, #319
Evidence: Dominion Schedule GS-4 filed 2025-12-09 effective 2026-01-01; Exhibit of Applicable Riders filed 2026-06-25 effective 2026-07-01; fifteen individual rider sheets; Virginia consumption tax and GS-4 sales-and-use surcharge; HB30 Item 3-5.24 for the data-centre tax.
Source
Corrected 2026-09-09 · defect #321 · field grid.carbon_intensity_g · British Columbia
| Withdrawn 2026-09-09 | Replacement |
|---|---|
basis claimed: average_operational_from_stated_mix basis recorded: none Published from initiation until 2026-09-09 | 26 gCO2eq/kWh operational_combustion modelled_from_mix derived from the market's own published fuel mix using carbon_factors.OPERATIONAL, which is cited and versioned |
The published figure was 12 gCO2eq/kWh. The market's own published fuel mix is 7% gas, and gas alone contributes 7% x 370 = 25.9 gCO2eq/kWh at the operational combustion factor this product cites. Nothing else in that mix can bring the number down: hydro, wind and solar are zero at this boundary, and biomass is zero by accounting convention rather than because nothing burns.
A figure BELOW the operational floor is impossible on any basis. Every alternative boundary -- life-cycle, CO2-equivalent -- is HIGHER than operational combustion, never lower, because each one adds terms. So this was not a boundary difference. The published number was simply not derived from the mix published beside it.
From initiation until 9 September 2026. It was detected by the product's own coherence check, which compares each published intensity against the floor implied by that market's own stated mix, and reported BELOW_ITS_OWN_FOSSIL_FLOOR.
26 gCO2eq/kWh, derived from the same published mix using the cited operational table. The value is now reproducible from published inputs by anyone, and the coherence check cannot contradict it by construction.
It is not a MEASURED figure. It remains a modelled estimate from a stated mix and keeps its existing decision-eligibility rank of insufficient_evidence. Deriving a number correctly is not the same as observing it, and the gain here is coherence rather than accuracy. An official published intensity from the operator, the regulator or the national inventory would supersede it.
The number went UP, from 12 to 26. The market looks slightly less clean than it did, which is the direction the evidence points.
None. carbon_intensity_g is not an input to the PGIQ Rating -- the carbon pillar is an analyst assessment -- so no tier, score, pillar or outlook moved.
16.712 gCO2eq/kWh · Environment and Climate Change Canada, National Inventory Report 1990-2024, Annex 7, Table A7-11, read directly from the published workbook · found 2026-09-09
generation intensity -- the inventory's own term for emissions per kWh delivered to the grid, which is this product's declared operational_combustion boundary
British Columbia generated 2.57% of its electricity from natural gas in 2024, not the 7% this record relied on. The operational floor was therefore 9.5, not 25.9, and the originally published 12 was ABOVE its floor rather than below it.
Wrong. The check that triggered it compares the published intensity against the published fuel mix and reports that they disagree; it cannot say which of the two is at fault, and here it was the mix. The number moved from 12 to 26 while the measured value is 16.7, so the correction moved it 9.3 further away.
The official figure is published on the market record as carbon_official_reference, with the served value named. The served value has not yet been changed: replacing it means changing carbon intensities on nine Canadian markets at once, which is a reviewed change, and this record will be superseded in full when that review concludes.
Because the alternative is a corrections page carrying a claim its author already knows to be false.
Defects closed: #321
Evidence: The market's own published fuel mix and carbon_factors.OPERATIONAL.
Methodology
Corrected 2026-09-09 · defect #322 · field grid.carbon_intensity_g · Ontario
| Withdrawn 2026-09-09 | Replacement |
|---|---|
basis claimed: average_operational_from_stated_mix basis recorded: none Published from initiation until 2026-09-09 | 37 gCO2eq/kWh operational_combustion modelled_from_mix derived from the market's own published fuel mix using carbon_factors.OPERATIONAL, which is cited and versioned |
The published figure was 31 gCO2eq/kWh against a mix of 10% gas, which contributes 10% x 370 = 37.0 on its own. Smaller in absolute terms than British Columbia, and the same defect: 6 gCO2eq/kWh is the entire margin between a defensible number and one the product's own coherence check rejects.
As for British Columbia: every alternative boundary is higher than operational combustion, so nothing explains a figure below the operational floor.
From initiation until 9 September 2026, detected by the same coherence check.
37 gCO2eq/kWh, derived from the same published mix using the cited operational table.
Not a measured figure. It remains a modelled estimate from a stated mix at the same eligibility rank, and an official published intensity would supersede it.
The number went UP, from 31 to 37.
None. carbon_intensity_g is not an input to the rating, so no tier, score, pillar or outlook moved.
64.909 gCO2eq/kWh · Environment and Climate Change Canada, National Inventory Report 1990-2024, Annex 7, Table A7-7, read directly from the published workbook · found 2026-09-09
generation intensity -- emissions per kWh delivered to the grid, this product's declared operational_combustion boundary
Ontario generated 15.52% of its electricity from natural gas in 2024, not the 10% this record relied on, so its operational floor was 57.4 rather than 37.0.
Right in direction and short in size. Both the withdrawn 31 and the replacement 37 sit below the floor Ontario's own generation implies, and the measured value is 64.9. Ontario genuinely had the defect this record describes; the corrected figure simply does not go far enough.
The official figure is published on the market record as carbon_official_reference, with the served value named. Changing the served value is a reviewed change covering nine Canadian markets together.
The same reason as British Columbia: a known error that a reader cannot see is worse than one that is stated.
Defects closed: #322
Evidence: The market's own published fuel mix and carbon_factors.OPERATIONAL.
Methodology
Build pgiq-20260926-224303.