Turn a forecast into a commitment.
Using wind-generation forecasts to support advance commitments to the electricity grid.
External case study
Original work: Google × DeepMind. Commentary by Meromorphic.
- Forecast
- Commit
- Evaluate
Problem
Variable wind generation makes it difficult to promise electricity delivery at a particular time. A forecast becomes commercially useful when it supports a decision about that commitment.
Result
Their 2019 account reports roughly 20% more value than a baseline with no time-based commitments. This is not 20% more electricity generated, nor a comparison against every alternative forecasting method.
Method
Google and DeepMind describe forecasting output 36 hours ahead using weather forecasts and historical turbine data, then recommending hourly delivery commitments a day in advance.
Why it matters
A generation forecast can improve when and how much power an operator commits to deliver. Measuring the value of those commitments connects prediction quality to a commercial decision.
Methodology and limitations
- This is a historical company account with limited public experimental detail; it is not an independently replicated result.
- Market rules, forecast uncertainty and imbalance penalties affect the value of delivery commitments.