WIND POWER PREDICTION BASED ON WORD EMBEDDING ATTENTION AND TEMPORAL CONVOLUTION

Considering the randomness of wind power variation, it is difficult to accurately predict wind power generation. In this context, this paper tentatively proposes a new time series model called the Attention-WE-TCN-LSTM. This model mainly included two parts: the TCN structure of word embedding and attention mechanism. By utilizing two different attention features for feature fusion through word embedding encoding, a certain predictive ability was achieved. Based on comparative experiments and ablation experiments, it not only demonstrates that the proposed model has certain advantages compared with traditional time series models, but also demonstrates its effectiveness.