勇 晔, 薛溟枫, 毛晓波. Agricultural load forecasting model based on multivariate temporal decoupling and multimodal learning[J]. 2025, 27(5). DOI: 10.3969/j.issn.1009-1831.2025.05.004.
As the agricultural load is greatly affected by meteorological factors and a single decomposition method cannot fully extract the multidimensional features existing between multiple inputs
an agricultural load forecasting model based on multivariate variational mode decomposition combined with SVR Bi GRU TCN combined model is proposed. Firstly
using multivariate variational mode decomposition to adaptively decompose historical agricultural loads and meteorological characteristics
real-time mining of modal components with different feature scales between data is carried out. Then
based on the inherent properties of each modal component
SVR
Bi-GRU
and TCN models are established to extract feature information at different time scales
thereby achieving accurate prediction of future 1-hour agricultural loads. The experimental results show that compared with the SVR model
Bi-GRU model
and TCN model
LSTM model and CNN-BiLSTM model
the proposed prediction model can effectively improve the prediction accuracy.