Data-Driven Physics Based Ultra-short-term Wind Power Prediction Model: A VMD-LSTM and WRF Based Approach

In this paper, an ultra-short-term wind power prediction model is proposed. First, based on the Weather Research and Forecast (WRF) model, the predicted wind speed of the wind power plant is obtained. Second, extract log history data from wind farm data acquisition and supervisory control system (SCADA). After, a variational mode decomposition (VMD) is applied to process historical wind speed of the wind farm to create 5 training data sets. Finally, Using long short-term memory (LSTM) network as a predictive model. The historical wind speed, historical wind power and WRF predicted wind speed are used as inputs to output the predicted power for the next 4 hours. For evaluating the prediction accuracy of this model, considering root mean square error (RMSE) and mean absolute error (MAE) as error evaluation metrics, the proposed model is compared with 3 other algorithms. The results show that LSTM has the best prediction performance.