Wind Power Prediction Based on Hybrid Neural Network and Wind Power Curve

In this paper, The traditional Recurrent Neural Network (RNN) can predict the time series data. However, there is always a certain deviation between the predicted value and the real value. In this paper, a hybrid neural network model consisting of attention mechanism (AM), convolutional neural network (CNN) and long and short memory network is constructed by using wind power curve. The experimental results demonstrate a significant improvement of the proposed model over traditional neural networks, particularly the traditional RNN network such as LSTM and CNN through the performance evaluation index parameters MAE and MSE, and the method has higher wind power prediction accuracy and smaller error value.