全球气候变暖已经成为人类迫切需要解决的难题,精确预测全球气温变化趋势对于把握气候发展状态、维护生态环境具有重要的意义。文章提出一种基于变分模态分解(VMD)与长短记忆神经网络(LSTM)的气温预测模型(VMD-LSTM),实现了对全球月平...全球气候变暖已经成为人类迫切需要解决的难题,精确预测全球气温变化趋势对于把握气候发展状态、维护生态环境具有重要的意义。文章提出一种基于变分模态分解(VMD)与长短记忆神经网络(LSTM)的气温预测模型(VMD-LSTM),实现了对全球月平均气温的准确预测。首先,对全球月平均气温进行VMD分解,得到了7个分量。其次,构造了LSTM一步预测模型对每一个VMD分量进行了预测。最后,根据VMD分量的预测值得到了全球平均气温的预测结果。数值实验中讨论了LSTM、VMD-LSTM、GRU与VMD-GRU四种预测模型,其中文章提出的VMD-LSTM的预测效果最好,其R2值为0.872,MAPE为0.664%,RMSE为0.121。实验结果表明,文章提出的VMD-LSTM预测模型能够有效预测气温。Global warming has become an urgent problem for mankind. Accurate prediction of global temperature is of great significance for grasping the state of climate development and maintaining the ecological environment. In this paper, a temperature prediction model (VMD-LSTM) is proposed based on the variational modal decomposition (VMD) and the long-short memory neural network (LSTM). This model was used to make accurate predictions of global average monthly temperatures. Firstly, the global average monthly temperature is decomposed into seven components by VMD. Second, The VMD component is predicted by the LSTM one-step prediction model. Finally, the prediction of global average temperature was obtained based on the forecasted values of the VMD components. The numerical experiment discussed four forecasting models: LSTM, VMD-LSTM, GRU, and VMD-GRU, among which the VMD-LSTM model proposed in the paper showed the best prediction performance, with an R2 value of 0.872, MAPE of 0.664%, and RMSE of 0.121. The experimental results indicate that the VMD-LSTM prediction model proposed in the paper can effectively predict temperature.展开更多
文摘全球气候变暖已经成为人类迫切需要解决的难题,精确预测全球气温变化趋势对于把握气候发展状态、维护生态环境具有重要的意义。文章提出一种基于变分模态分解(VMD)与长短记忆神经网络(LSTM)的气温预测模型(VMD-LSTM),实现了对全球月平均气温的准确预测。首先,对全球月平均气温进行VMD分解,得到了7个分量。其次,构造了LSTM一步预测模型对每一个VMD分量进行了预测。最后,根据VMD分量的预测值得到了全球平均气温的预测结果。数值实验中讨论了LSTM、VMD-LSTM、GRU与VMD-GRU四种预测模型,其中文章提出的VMD-LSTM的预测效果最好,其R2值为0.872,MAPE为0.664%,RMSE为0.121。实验结果表明,文章提出的VMD-LSTM预测模型能够有效预测气温。Global warming has become an urgent problem for mankind. Accurate prediction of global temperature is of great significance for grasping the state of climate development and maintaining the ecological environment. In this paper, a temperature prediction model (VMD-LSTM) is proposed based on the variational modal decomposition (VMD) and the long-short memory neural network (LSTM). This model was used to make accurate predictions of global average monthly temperatures. Firstly, the global average monthly temperature is decomposed into seven components by VMD. Second, The VMD component is predicted by the LSTM one-step prediction model. Finally, the prediction of global average temperature was obtained based on the forecasted values of the VMD components. The numerical experiment discussed four forecasting models: LSTM, VMD-LSTM, GRU, and VMD-GRU, among which the VMD-LSTM model proposed in the paper showed the best prediction performance, with an R2 value of 0.872, MAPE of 0.664%, and RMSE of 0.121. The experimental results indicate that the VMD-LSTM prediction model proposed in the paper can effectively predict temperature.
