Long term¶
The long-term quantities of the waterfall are the analysis window recomputed with the long-term wind climate of the site, through a representative period. Year-to-year variability is removed; seasonal and diurnal cycles are preserved. This separates the effect of a low-wind period from underperformance.
Annual production depends primarily on the wind resource of the year. A comparison between two periods, or between two wind farms, therefore requires a common climate reference. Tipspeed builds this reference and expresses results on two periods, linked by an explicit windiness step:
- the long term: the analysis window under the long-term climate, built as a representative period;
- the current period: the actual analysis window.
Long-term wind reconstruction¶
A machine learning model correlates an ensemble of multi-year reanalysis and mesoscale sources, including ERA5, with the inflow of the site, as measured on site and as simulated by the wind flow model. The model is trained over a period in which both are available. It then reconstructs the site inflow over the full history of the reanalysis sources, typically more than fifteen years.
Representative period¶
The representative period is assembled from the long-term history. For each calendar day of the analysis window, one source day is drawn at random among all the years available for the same calendar day. Its 10-minute data is placed on the analysis calendar. Year-to-year variability is removed, while the seasonal and diurnal cycles of the site are preserved.
The set of candidate days and the output window are independent. A climatology built from complete calendar years can therefore be projected onto an analysis window that extends beyond those years.
The representative inflow is then processed by the same wake and turbine performance models as the rest of the digital twin. The result is the Expected (long term) time series: the expected production of an optimised wind farm under climate conditions representative of the long term (see Timeseries models).
Use in the energy chain¶
The energy waterfall starts on the long-term climate, at Expected gross (long term). It reaches the current period through a windiness step, equal to the difference between the expected gross production over the long term and over the current period. Losses also depend on the wind resource of the period. Site-specific performance losses and wake losses therefore each include their own windiness correction, equal to the difference between the value of that loss over the current period and over the long term.
In the resulting chain, windiness is isolated in a single, explicit step, and each remaining step represents a performance or availability effect. The waterfall labels the quantities computed over the representative period "(long term)". The definitions and formulas of each step are given in the Metrics glossary.