Forecdiction of Short-term Wind Power in Ice Season Based on Wind Turbine Operating State and Deep Learning

Wind power has strong volatility and randomness. In the ice season, wind power is more susceptible to the climatic conditions, which is difficult to predict, and has an impact on grid scheduling. For this problem, a wind power prediction method based on wind turbine operating state and deep learning. This method takes the historical meteorological data and wind turbine power as input, and establishes a wind turbine operation state prediction model through a fully connected neural network to realize wind turbine operation state prediction. The CNN-BiLSTM model is used to predict the wind power without the ice, and then convert the power according to the running status of the turbine to forecasting the electrical power of the wind during the ice. The experimental findings indicate that the proposed model has good prediction performance in the ice season and verify its accuracy by comparing with the traditional method.