Accuracy of a short-term wind power forecasting model based on deep learning using LiDAR-SCADA integration: A case study of the 400-MW Anholt offshore wind farm

Countries are increasingly investing in various renewable energy facilities to meet their national greenhouse gas reduction targets. This has led to an increase in highly variable renewable energy generation within the power system, highlighting its poor stability. To solve this issue, one of the measures required is an effective forecasting technique. This study targeted a 400-MW offshore wind farm, examining the diverse uses and effects of different input features with a short-term wind power forecasting model. Using Laser Imaging Detection and Ranging (LiDAR) and Supervisory Control and Data Acquisition (SCADA) data as well as deep-learning algorithms, we developed a 3-h wind power forecasting model. An optimal model was subsequently selected by evaluating diverse deep-learning algorithms and combinations of input feature groups. The resulting model had an average prediction accuracy of 6.8% and an output error ranging 2.11–10.95% within the 3-h prediction interval. Finally, a forecasting model with integrated LiDAR and SCADA data and using a stacking recurrent layer was confirmed to be able to limit prediction errors more effectively than other models. This is because offshore LiDAR data provide information on the external environment, while SCADA data provide dynamic information on the mechanical factors of the wind turbine. This integrated approach enables the effective modeling of interactions between external environments and wind turbines, further improving the robustness and generalization ability of forecasting models. The results of this study are expected to support the higher performance and stable operation of wind power forecasting models, while providing useful information for their application in actual offshore environments.