Robust Short-Term Wind Power Forecasting using a Multivariate Input and Hybrid Architecture

Wind power is a clean, efficient and sustainable source of electricity that is highly regarded in the renewable energy industry. However, wind power is significantly dependent on weather, especially wind speed. Therefore, making an accurate forecast of the generating capacity of a wind power plant is very important to help manage and optimize the operation of the power generation system. This paper proposed wind power forecasting models using time series forecasting methods such as LSTM, CNN and the combination of CNN-LSTM with multivariable and univariate input. These models are evaluated with a dataset collected from wind power plant BT2 - Quang Binh. The results show that the combined model of CNN-LSTM with multivariable input gives more accurate predictive results than other proposed models in the Turbine 27 dataset in term of Root Mean Square Error and Mean Absolute Percentage Error.