Fig 1.
Flowchart of VMD optimization in SSA.
Fig 2.
Flow of wavelet threshold denoising.
Fig 3.
Network structure of BiLSTM.
Fig 4.
Operation procedure of CNN-BiLSTM model.
Fig 5.
Flow of SSA-VMD-WST model.
Fig 6.
The 13 IMF components obtained by SSA-VMD decomposition of the original signal.
Table 1.
Data decomposition error under different parameters.
Table 2.
Correlation between the IMF component and the original signal.
Fig 7.
SNR under different wavelet bases, wavelet thresholds and decomposition layers.
Fig 8.
Data fluctuations under different denoising methods.
Table 3.
Comparison of index under different denoising methods.
Fig 9.
CNN-BiLSTM trend fitting of maximum temperature under different denoising methods.
Table 4.
Comparison of prediction results.