Using Probabilistic Energy Production Forecasts to Dynamically Estimate Forecast Accuracy

Renewable energy forecasting is important to participants within power markets,
including those that operate assets, bid into power markets, and operators who
delegate secondary power production. Weather forecasts have variable accuracy from
day to day, which means that any operations that use weather forecasts to estimate
energy production are at risk of volatility. Fortunately, probabilistic weather and energy
production forecasts offer a way for users to assess the confidence and expected error
of any given forecast.

This presentation will highlight a method for identifying periods when renewable energy
forecasts are particularly confident and most likely to be accurate. We use probabilistic
forecasts to identify when a distribution is confident, then provide a method for
estimating the amount of error in energy production estimates that will occur when using
renewable energy forecasts in an operational setting. We also present findings of when
and where renewable energy forecasts are the most confident, which also highlights
which regions may find the most benefit from relying on renewable energy production
forecasts.

These methods can be used to inform risk within systems that rely on renewable energy
production, either through automation or through manual monitoring of forecast
confidence levels. We focus on subseasonal to seasonal timescales in this
presentation, while the methodology presented can be applied to forecasts for any time
cadence.