I. Short-term forecasting of offshore wind farm production. Developments of the Anemos project.

Due to the large dimensions of offshore wind farms, their electricity production must be known well in advance to allow an efficient integration of wind energy into the European electricity grid. For this purpose short-term wind power prediction systems which are already in operation for onshore sites have to be adapted to offshore conditions, which has been a major objective of the EU-project ANEMOS. The paper presents the offshore results of the project partners in a cumulative way. In general, it has been found that the accuracies of wind speed predictions for the offshore sites Horns Rev and FiNO1 are similar or better than for single onshore sites considering that the mean producible power is twice as high as onshore. A weighted combination of two forecast sources leads to reduced errors. A regional forecast of the aggregated power output of all projected sites in the German Bight with a total capacity of 25 GW benefits from spatial smoothing effects by an error reduction factor of 0.73, showing an RMSE of 3GW. An aggregated forecast for the sum of on- and offshore production in Germany with a total capacity of 50GW would benefit from an error reduction factor of 0.43, leading to an RMSE of 3.5 GW. The project partners also investigated the most important parameters which influence the wind speed profile offshore. A new air-sea-interaction model for calculating marine wind speed profiles was developed, i.e. the theory of inertially coupled wind profiles (ICWP). Evaluation with Horns Rev and FINO1 data showed good agreement, especially regarding wind shears. Next, emphasis was given on modelling spatio-temporal characteristics in large offshore farms. New approaches were developed to model wakes behind such farms. Wake losses are anticipated to be at least 5-10% of power output. Wind speed recovery can be predicted to occur between 2 and 15 km downwind of such farms according to the model type chosen. Also, a comparison of mesoscale model results with WAsP predictions was performed to quantify gradients of wind speed over large wind farms. Moreover, the contribution of satellite data in offshore prediction was studied. For the complex situation in the Strait of Gibraltar, a semi-empirical model was developed. Finally, various physical and statistical (i.e. neural networks) models were calibrated on power data from two offshore wind farms: Tunoe and Middelgrunden in Denmark.