Wind power Prediction based on the fusion of CN N-GRU combined neural network and attention mechanism

Abstract

Accurate prediction of wind power plays an i mportant role in long-term stable and safe operation of pow er system. Wind power generation can save a lot of energy a nd solve the problem of high cost and low utilization rate of traditional thermal power generation. In this paper, a comb ination of CNN-GRU-Attention model is proposed, and thre e different data decomposition methods are used for data pr e-processing, and the optimal data pre-processing method is selected for the model. Meanwhile, the fusion model is comp ared with CNN-LSTM-ATTENTION, GRU, RNN and other traditional single models. By analyzing MAE, RSME, R an d other evaluation indexes, it is concluded that the predictio n accuracy of CNN-GRU-Attention model is the best. In wi nd power prediction, the predicted value of the model is clos est to the actual value, and the prediction effect is higher tha n LSTM, BPNN and other traditional networks, which is of great significance for wind turbine power prediction.