Aiming at the problems that traditional wind power forecasting methods have few considerations in terms of the close relationship between time series and spatial global features and parallel processing
and the prediction reliability is limited
an ultra-short term wind power forecasting method based on cross-attention fusion of spatial-temporal features of TCN-SENet-Transformer is proposed. Firstly
squeeze and excitation networks(SENet) are used to adjust the channel feature weights
and temporal convolutional networks(TCN) are used to capture the spatial features of the data. Meanwhile
Transformer is used to identify the long-term timing characteristics of multi-feature data. Then
cross-attention(CA) is introduced to integrate temporal and spatial features. Finally
the actual data of a wind farm in China is used to forecast ultra-short term wind power
and the comparison is made with other prediction models. The results of example analysis show that the proposed combined prediction model effectively improves the prediction accuracy.
LE GOFF LATIMIER R, LE BOUEDEC E, MONBET V. Markov switching autoregressive modeling of wind power forecast errors[J]. Electric power systems research, 2020, 189: 106641.
LIU M D, DING L, BAI Y L.Application of hybrid model based on empirical mode decomposition, novel recurrent neural networks and the ARIMA to wind speed prediction[J]. Energy conversion and management, 2021, 233: 113917.
WANG K, ZHANG Y, LIN F, et al.Nonparametric probabilistic forecasting for wind power generation using quadratic spline quantile function and autoregressive recurrent neural network[J]. IEEE transactions on sustainable energy, 2022, 13(4): 1930-1943.
BAI S J, ZICO KOLTER J, KOLTUN VLADLEN. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling[J]. arXiv, 2018, 1803.01271: 3-4.
CHEN Y B, XU J J.Solar and wind power data from the Chinese state grid renewable energy generation forecasting competition[J]. Scientific data, 2022, 9: 577.