To address the limited accuracy of multi-step wind speed predictions using conventional methods
this paper proposes a hybrid model that integrates Singular Spectrum Analysis (SSA) and Variational Mode Decomposition (VMD) with a Time Convolutional Network (TCN)-Informer architecture. Firstly
SSA is used to suppress noise in the original wind speed data and reduce its instability. Next
VMD is employed to reduce the complexity of the wind speed sequence
with each component then input into TCN's feature extraction module to capture temporal features and enhance local information representation. Finally
by fusing temporal and spatial features from each modal component and inputting them into Informer's self-attention model
long-term dependence relationships are modeled to obtain multi-step wind speed predictions. The proposed model was validated using measured wind speed data from a meteorological tower at a wind farm in Yunnan province
China. The results show that the MAPE for 6-step and 12-step predictions were only 1.63% and 2.25%
respectively
demonstrating significantly accuracy in short-term multi-step wind speed prediction.
VASWANI A,SHAZEER N,PARMAR N,et al.Attention is all you need[C]//Proceedings of the 31st International Conference on Neural information Processing Systems. Long Beach, USA, Curran Associates Inc, 2017: 6000-6010.
ZHOU H Y, ZHANG S H, PENG J Q, et al.Informer: beyond efficient transformer for long sequence time-series forecasting[J]. Proceedings of the AAAI conference on artificial intelligence, 2021, 35(12): 11106-11115.
DU W H, ZHOU J, WANG Z J, et al.Application of improved singular spectrum decomposition method for composite fault diagnosis of gear boxes[J]. Sensors, 2018, 18(11): 3804.