Posters

Abstract

Climatology is commonly used in the wind industry to compute production for the following year. It has been found that using monthly climatology averages is not effective for anomalous cases because monthly wind speed anomaly predictions change every month. Therefore, some annual predictions must be updated monthly or quarterly to fulfil predicted budgets calculated with historical data. A new methodology for seasonal forecasting has been developed to improve on climatology for 6-month wind speed anomaly predictions. This methodology combines 30-year historical time series and seasonal climate models with statistical methods and machine learning techniques, which does not require prior wind-speed measurements. A modelled 30-year time series from the Weather Research and Forecast model (WRF) at 3km resolution prepared by Vortex is used as a reference for the climate period 1991-2020. Seasonal climate models from Copernicus Climate Change Service (C3S) are used for 6-month projections. The seasonal models used are from European Centre for Medium-Range Weather Forecasts (ECMWF), the UK Met Office, Météo-France, the German Weather Service (Deutscher Wetterdienst, DWD), the Euro-Mediterranean Center on Climate Change, the US National Weather Service's, and Japan Meteorological Agency (JMA). All the gathered information is analysed on a site-specific basis using intelligent analysis to evaluate the best performance for each exact location. Seasonal single and post-processed models, multi-models, statistical projections, persistence and climatology are all studied to provide the best anomaly prediction for each point. Furthermore, budgets for Annual Energy Production (AEP) can be updated every month. And, better information can be produced for planning wind farm operations, maintenance, storage and construction for the months ahead.