A combined short-term wind speed forecasting model based on CNN–RNN and linear regression optimization considering error.

Wind speed forecasting is the key to wind power conversion and management in smart grids. In this paper, a new hybrid model is proposed, which is composed of empirical mode decomposition, a convolutional neural network, a recurrent neural network and a linear regression network considering the model error. In this model, empirical mode decomposition is used to decompose the original wind speed series into multiple subseries, five types of neural networks are used as predictors for each subseries, and a linear regression network is used as the second-level predictor to forecast the wind speed and error series of each model. To verify the prediction ability of the model, experiments are performed using the wind speed data of three stations in the actual wind farm for four months. The results show that the proposed hybrid prediction model has better accuracy and stability than any single neural network model and any neural network model with decomposition preprocessing, which also shows that combination forecasting is a robust and reliable wind speed forecasting method. • Decomposition-based hybrid wind speed forecasting model using CNN–RNN. • CNN is used to extract multi-scale features of wind speed. • Linear regression network is used to combine forecasting while considering error. • The accuracy and stability of the model are evaluated on the wind farm datasets. [ABSTRACT FROM AUTHOR]