Forecasting Maximum Wind Speed at Offshore Sites

Abstract: For energy companies involved in the construction and operation of offshore wind power plants, such as Vattenfall, the knowledge of maximum wind speed is critical for logistics, safety and economic reasons. This thesis investigates the possibility to forecast maximum wind speeds at offshore sites, studies the accuracy of these forecasts, and details the employed methodology, so that it can be adapted to other cases in the future. In order to produce maximum wind speed forecasts, different statistical models were selected, some of them appropriate for short-term predictions (from 1h to 6h ahead), the others aiming at longer-term predictions (days ahead, up to 72h). The methodology consisted in selecting the right parameters for each model, depending on wind measurements and weather forecasts at the tested sites. Then forecasts were issued using the models’ equations, forecasted maximum wind speeds were compared with the real values, and the model delivering the best forecasts selected. The study demonstrated that, if appropriate statistical models were chosen – such as Vector Auto-Regression for short-term, and Generalized Additive Model for long-term – the average errors of precision for maximum wind speeds prediction were lower than 2 m/s, making the forecasts accurate enough to be used. Some work on the models still has to be done before they can be fully integrated into Vattenfall’s in-house weather forecasting system, but the first results are promising.