Attention-based convolutional neural network-long short-term memory network wind power forecasting

To enhance the precision of wind power forecasting, this study proposes an attention-based convolutional neural network-long short-term memory (AT-CNN-LSTM) prediction model. The model is designed to address the intermittent and sequential characteristics of wind power data. Initially, a CNN is employed to extract multidimensional features from the dynamic changes observed in wind power data. Subsequently, the extracted feature vectors are transformed into a sequential format and utilized as input for the LSTM network. To optimize the model, an attention mechanism is incorporated, assigning varying weights to the hidden layers of the LSTM network. This approach enhances the influence of crucial information, thereby improving wind power prediction. Experimental results conducted on wind power data obtained from a local wind farm demonstrate that the proposed model surpasses both LSTM and CNN-LSTM models in terms of prediction accuracy.