When a planned U.S. power plant drops out of EIA's monthly survey, it rarely comes back under a new name

Description and use

Only 2.35% of 'not online within a year' capacity in EIA-860M was re-filed under new generator ids at the same plant and later came online (2.52% with broader matching). Published delivery rates barely move (32.47% to 32.49%). The ten largest links all verified as one project under hand-reading.

Problem

How much of the 'never came online' capacity in EIA-860M is actually a re-filing under new generator ids at the same plant, rather than a true cancellation?

Outcome

Of the capacity scored 'not online within a year of its planned month', only 2.35% was re-filed at the same plant under new ids that did come online (2.52% if counting any later generator of the same fuel). The published delivery rates move from 32.47% to 32.49%, with no breakdown rising by more than 0.3 points. The ten largest links all read as one project under hand-verification.

Workflow

• Score each planned generator in EIA-860M for whether it came online within one year of its planned month.
• Identify plants where a generator left the list without operating and later re-filed under new generator ids at the same plant.
• Compare the share of 'not online within a year' capacity that is re-filed vs. truly cancelled.
• Check whether loosening the rule (counting any later generator of the same fuel at the plant) changes the estimate.
• Verify delivery rates by technology, stage, and lead time to see if re-filing materially affects published rates.
• Hand-read the ten largest links against three fixed rules to confirm they represent one project.
• Report the re-filing rate, delivery rate impact, and link verification results.

Evidence

Of the capacity scored 'not online within a year of its planned month', 2.35% was re-filed at the same plant under new ids that did come online. We predicted 3% to 20%; that failed, low.

Loosening the rule does not change that: counting any later generator of the same fuel at the plant that ever came online gives 2.52%.

The share of planned capacity online by its planned month moves from 32.47% to 32.49%, and in no breakdown by technology, stage or lead time does it rise by more than 0.3 points.

Three plants carry 92.4% of the re-filed capacity, and the whole combined-cycle effect is one Ohio plant, re-filed as four generators.

One Texas plant came back 20 months late, so a re-filing is not the same as an on-time plant.

The ten largest links were read by hand against three rules fixed in advance; all ten read as one project. One smaller link does not (an existing unit added to the list) and is reported as such.

We found no published crosswalk or estimate for re-filed planned generators in EIA-860M; PUDL records one plant whose ids changed between the monthly and annual surveys.