Google and Google DeepMind
Mixed evidenceContinuous decisioningStep before the decisionThat variable wind generation cannot be committed to the grid on a useful day-ahead schedule.
Logistics · Central United States
Executive brief
Prediction mattered because it was connected to a changed commitment decision; forecast accuracy alone would not create grid value.
AI value · Value of wind energy versus no time-based commitments
Company-reportedRoughly 20% higher value
Google-reported value metric; monetary definition, forecast error distribution, and independent audit were not published.
Before
Estimate uncertain output without the ML forecast and avoid firm hourly commitments. → Treat wind as less schedulable and therefore less valuable.
After
Predicts wind output 36 hours ahead from weather forecasts and turbine history. → Recommend and submit hourly delivery commitments a day ahead.
Human boundary
The model recommends commitments; accountable operators and market processes determine submitted schedules.
Why it matters
That variable wind generation cannot be committed to the grid on a useful day-ahead schedule.
Before
How the work ran before the change.
Wind operator
Estimate uncertain output without the ML forecast and avoid firm hourly commitments.
ControlNo time-based commitment baseline
Grid
Treat wind as less schedulable and therefore less valuable.
ControlGrid balancing rules
What changed
That variable wind generation cannot be committed to the grid on a useful day-ahead schedule.
Decision rightSelection moves from a fixed rule to the model
After
How the same work runs now.
Neural forecasting model
Predicts wind output 36 hours ahead from weather forecasts and turbine history.
ControlForecast uncertainty
Optimization system and operator
Recommend and submit hourly delivery commitments a day ahead.
ControlOperator and market rules govern commitments
Exception path
Operators update positions and manage imbalance when weather or turbine availability diverges from forecast.
Work removed
Decision authority
The model recommends commitments; accountable operators and market processes determine submitted schedules.
Before
Wind operator
Estimate uncertain output without the ML forecast and avoid firm hourly commitments.
Control: No time-based commitment baseline
Grid
Treat wind as less schedulable and therefore less valuable.
Control: Grid balancing rules
After
Neural forecasting model
Predicts wind output 36 hours ahead from weather forecasts and turbine history.
Control: Forecast uncertainty
Optimization system and operator
Recommend and submit hourly delivery commitments a day ahead.
Control: Operator and market rules govern commitments
Work that left the path
Human role before
Operators managed wind output without a machine-generated 36-hour production and commitment recommendation.
Human role after
Operators oversee forecast-driven commitments and retain market and operational accountability.
AI roleNeural forecasting plus optimization that predicts output and recommends hourly grid commitments.
No time-based delivery commitments to the grid→Roughly 20% higher value
Results through 2019-02-26 after deployment beginning in 2018 · 700 MW of Google wind capacity in the central United States
Google-reported value metric; monetary definition, forecast error distribution, and independent audit were not published.
Anti-pattern
Reporting 20% more energy generated; the source says value, not production.
Questions
Portability conditions
Reputation risk
medium
Evidence and authority
Current · updated
1 independent, 1 primary; publication outcomes are verified and reported.
Bundle 1.0.0 · reviewed 2026-08-23 · stable ID 63b51d1a931c9f58
Google DeepMind and EMBL-EBI · Continuous decisioning
Researchers can use precomputed structure predictions, then decide what experimental validation is required.
Google Research Flood Hub · Continuous decisioning
A global model can transfer hydrology into ungauged basins.
Mayo Clinic primary-care practices · Continuous decisioning
A routine ECG can trigger evaluation for asymptomatic low ejection fraction before specialist suspicion.
Google · Continuous decisioning
Novel breakages need not wait for a comprehensive post-submit test cycle.