Illusionary wind energy predictability imperils the strategy of power grid decarbonization: Evidence from offshore Great Britain

A fundamental strategy in the energy transition away from fossil fuels is increased reliance on wind energy in the electric power sector, coupled with increased electrification throughout the economy. One of the challenges in implementing this strategy is achieving sufficiently accurate wind energy forecasts to maintain power grid stability. Given volatility and scale considerations, this challenge is particularly severe in the marine environment. Unfortunately, the forecast error metric devised and actively promoted by wind energy forecasters can strongly suggest that a forecast is accurate even when its forecast skill score is negative. As a result, the use of this metric creates an illusion of predictability. Because of this illusion, the challenge of “keeping the lights on” will become more difficult in Great Britain as large, planned increases in offshore wind capacity are achieved. To help dispel this illusion of predictability and the challenge it poses to the resiliency of the power grid and the energy transition, this paper presents a statistical method to improve the predictability of offshore wind energy generation using data from the British power grid. The methodology employed is a time-series approach called ARCH/ARMAX, which effectively models wind energy generation. Predictions from this model, made for an out-of-sample period, are significantly more accurate than those from current methods over the same timeframe, indicating that the resilience of the electricity system does not need to decline with greater reliance on offshore wind energy. However, the difficulty of dispelling the current illusion of predictability may mean that the power grid's resilience will be increasingly imperiled as the energy transition proceeds, unless a more robust regulatory process is adopted. • Poor predictability of wind energy generation imperils the stability of the power grid. • Normalizing the errors by capacity can create an illusion of predictability. • Improved recognition of forecasted meteorological conditions can improve matters. • Time-series modeling can also improve predictability.