Can we improve short-term wind power forecasts using turbine-level data? A case study in Ireland

摘要

Short-term wind power forecasting plays an important role for practitioners in the energy industry, as they are used to facilitate the decision-making process in the energy trading market. Despite the extensive literature existing for short-term wind power forecasting, there is still room for innovation in terms of data. At the present time, more and more near real-time data from different sources are available in wind farms, which can be fed to wind power forecasting models to improve the accuracy of wind power predictions for wind farms. In particular, large quantities of high resolution data collected at a turbine level are available for forecasters to train their models, which are often disregarded even if they can lead to a higher forecasting performance.The benefits of these type of data are shown with a case study using high resolution wind power generation data collected at a turbine level from two Irish wind farms, in which the forecast accuracy of a machine learning based forecasting model is analyzed using data at a turbine- and wind farm-level for time horizons up to 8-h ahead. The results show that the use of turbine-level data can increase the prediction accuracy for short-term wind power forecasts.