ValidAnytime
  • Try it free
  • Observatory
  • Pricing
  • Docs
Sign inStart free
ValidAnytime

A guaranteed false-alarm budget across your whole fleet, valid no matter how often you look.

Made by Compiled Intelligence — a frontier AI lab working on quantitative finance from first principles; ValidAnytime is the monitoring we built for our own model fleets, productized.

Product

  • How it works
  • Pricing
  • Observatory
  • Try it free
  • Dashboard

Solutions

  • Who it's for
  • For LLM engineers
  • For MLOps engineers
  • For SRE & platform
  • LLM eval regression
  • Model drift detection
  • Alert fatigue

Developers

  • Docs
  • Quickstart
  • Python SDK
  • API reference

Learn

  • Benchmark
  • Detector guides
  • Glossary
  • Compare
  • Why dashboards lie
  • Blog
  • Anytime-valid 101
  • Winter Storm Uri case study
  • Incident ledger
  • Observatory methodology

Company

  • Contact
  • Privacy
  • Terms

Observatory alarms & product updates, by email

Rare emails only. We store your address and the tag “footer” — nothing else. Unsubscribe anytime.

© 2026 ValidAnytime. All rights reserved.

Every alarm ships with its guarantee_tag and theorem_ref.

Observatory
TV

Has Tennessee Valley Authority's demand forecast drifted?

TVA · updated Jul 6, 22:08 UTC

No regime change detected

Tennessee Valley Authority — no forecast-error regime change detected

The day-ahead forecast's error has stayed within its calibrated band. Not a certificate that nothing happened.

quiet streak 5 monthsbudget ≈1 false page per BA-yearwatching since Jan 2024
464.05%
mean abs. forecast error

of demand (MAPE)

19,688 MW
mean demand
323 MW
typical error

median |error|

2,240 MW
worst-hour error

99th pctile |error|

The evidence, last two weeks

Because the baseline is the operator’s own forecast, seasonality is subtracted for free — what’s left is the miss. The detector pages only when that miss enters a new regime and clears the page line.

0 = the operator’s forecast is exactly rightforecast error — actual demand minus day-ahead forecast (MW)page line (≈1 false page per grid-year)accumulated evidence (anytime-valid log-statistic)Jun 22, 2026Jul 6, 2026
Real EIA-930 data for Tennessee Valley Authority — the last 336 hourly points. The forecast is the grid operator’s own day-ahead demand forecast (not ours); the error is actual − forecast, peaking at +4,880 MW in this window. The evidence statistic stayed under the page line: no regime change detected. That is not proof nothing happened. Back-in-band is a display heuristic, not a certificate of absence.

Forecast-error regime changes4 in the monitored history

Each is a dated page inside the stated budget. Labels attach only where a window overlaps a documented grid-stress event — never invented. We detect that the forecast’s error changed regime, not that the grid failed or why.

  • EndedJan 26, 2026 → Jan 28, 2026January–February 2026 cold wave (Winter Storm Fern)
    peak evidence
    9.30
    forecast miss at onset
    −2,837 MW
    load at onset
    97th pctile of this grid
    back in-band
    Jan 28, 2026
  • EndedJan 26, 2026 → Jan 28, 2026January–February 2026 cold wave (Winter Storm Fern)
    peak evidence
    10.06
    forecast miss at onset
    −4,468 MW
    load at onset
    89th pctile of this grid
    back in-band
    Jan 28, 2026
  • EndedFeb 21, 2025 → Feb 23, 2025
    peak evidence
    10.10
    forecast miss at onset
    −2,387 MW
    load at onset
    100th pctile of this grid
    back in-band
    Feb 23, 2025
  • EndedJan 10, 2025 → Jan 13, 2025
    peak evidence
    13.27
    forecast miss at onset
    −2,055 MW
    load at onset
    92th pctile of this grid
    back in-band
    Jan 13, 2025

How this verdict is formed

The hourly error runs through a conformal band that learns Tennessee Valley Authority’s own normal forecast-error rate, into a Shiryaev–Roberts e-detector on a bounded rolling window. An alarm fires only when accumulated evidence crosses the page line — and because the statistic is anytime-valid, checking as often as you like never inflates the false-alarm rate. Full recipe, and exactly what we do not claim, on the methodology page.

Embed this badge

A live status pill for Tennessee Valley Authority, refreshed every publish. Drop it in a README, status page, or dashboard — it always shows the current verdict and links back here. It reads no regime change while the forecast holds, and flips to forecast drift detected when the e-detector pages.

Tennessee Valley Authority demand-forecast monitor · ValidAnytimelive preview
Markdown
[![Tennessee Valley Authority demand-forecast monitor · ValidAnytime](https://validanytime.com/observatory/badge/TVA)](https://validanytime.com/observatory/TVA)
HTML
<a href="https://validanytime.com/observatory/TVA"><img src="https://validanytime.com/observatory/badge/TVA" alt="Tennessee Valley Authority demand-forecast monitor · ValidAnytime"></a>
Monitor your metricsBacktest your data