Combined forecasting model of sea surface temperature based on improved variational mode decomposition and gate recurrent unit
Lingzhi Tan, Guohui Li · 2024
Aiming at the non-stationary and non-linear characteristics of sea surface temperature, a combined forecasting model of sea surface temperature based on improved variational mode decomposition and gate recurrent unit is proposed. Aiming at the problem that the decomposition level and penalty factor of variational mode decomposition (VMD) must be set manually, VMD based on snow ablation optimizer (SAO) algorithm is proposed, named SAO-VMD. Aiming at the problem that the loss function of gate recurrent unit (GRU) will affect its accuracy and convergence speed, GRU based on improved loss function (ILF) is proposed, named as ILF-GRU. To begin with, SAO-VMD is used to decompose sea surface temperature data, and a series of intrinsic mode functions (IMFs) are obtained. Then, ILF-GRU is used to predict IMFs. Finally, the component prediction results are reconstructed to obtain the final prediction results. To prove the superiority of the proposed model, the average sea surface temperature of each point in designated areas $(5 \mathrm{~N} \sim 25 \mathrm{~N}$ and $110 \mathrm{E} \sim 170 \mathrm{E}$) from 1982 to 1992 is selected for simulation experiment. The experimental result shows that the proposed model is superior to all other comparative models, proving that it has higher prediction accuracy and performance.