Unfortunately, traditional risk assessment methods take neither the extremes of interannual variability in wind speeds nor the risk of global stilling into account. Rather, these methods base production forecasts on retrospective studies in which a single year of direct measurements is used to perform bias correction on reanalysis model data. Therefore, investors and planners often do not have tools to quantify either the actual potential extremes or the impact of trends on wind energy resources. Both issues introduce risk that future wind speeds will fall outside of the limited historical record.
In this presentation, we propose a more comprehensive risk assessment method that instead utilizes a forward looking ensemble of stochastic weather simulations to reflect climate trends and incorporate interannual variability. This stochastic simulation approach allows planners to construct an arbitrarily precise distribution of outcomes and better prepare for extremes.
To illustrate these methods in practice, we selected a location in central Kansas and produced 1000 hourly stochastic weather simulations of the year 2023. For each simulation, we estimated wind energy production and compared the distribution of production outcomes to observations. We also estimated hypothetical wind energy production from the historic record and found that while the 2023 extremes are outside the range of the historic data (1950-2022), the stochastic method is able to recreate events as extreme (or more extreme) than the observed deficit in 10% of simulations.
We further simulated the 15 year period from 2022 though 2036 for 100 representative utility scale wind farms across the United States. Examining theoretical uncertainty in results, we found estimates from stochastic simulations to be more robust and less prone to sampling noise. Importantly, we discovered regional trends in wind strength and direction with profound effects on the resulting power generation estimates that compound over time – emphasizing the need for local trend-corrected ensemble based analyses for wind power generation.

