This study improves wind power forecasting accuracy with machine learning. A hybrid approach of Empirical Mode Decomposition (EMD) and Support Vector Regression (SVR), with a hyperparameter tuning, leads to higher accuracy. The wind turbine power data used and found that EMD-based approach is better than SVR. Applying parameter tuning showed significant improvement in wind power forecasting. The study's findings are important for energy traders looking to maximize profits in the intraday market.

An important design criterion for the electric drive system of a wind turbine is the fulfilment of grid codes given by transmission system operators. The grid codes state how wind turbines/farms must behave when connected to the grid in normal and abnormal conditions. A type of testing equipment that comprises the use of fully-rated voltage source converter in back-to-back configuration for grid code testing is proposed. Test results of a 4 MW wind turbine and an 8 MW test equipment are shown.