基于VMD-TCN的月降水量预测模型

Monthly Precipitation Prediction Model Based on VMD-TCN

准确的月降水量预报对水资源的合理开发利用及洪涝灾害的预测和防治具有重要意义。针对传统降水预测方法信息挖掘能力不足的问题,提出了一种基于VMD--TCN的月降水量组合预测模型。该模型使用VMD将原始序列分解为一系列相对平稳的子序列,然后利用TCN对各子序列分别进行预测,输出各子序列预测结果,叠加得到最终预测结果。将VMD-TCN降水预测模型与EMD-TCN模型和VMD-LSTM模型进行对比。结果表明:VMD-TCN模型的R^(2)可达0.98,与EMD-TCN和VMD-LSTM模型相比,RMSE分别减少了83.85%和43.56%,MAE分别减少了84.25%和43.60%,明显优于EMD-TCN和VMD-LSTM。在精度高于VMD-LSTM模型的基础上,VMD-TCN模型基于卷积的并行思想,运行速度是VMD-LSTM模型的2倍多,为月降水量预测提供了一种有效方法。

Accurate monthly precipitation forecast is of great significance to the rational development and utilization of water resources as well as the prediction and prevention of flood and waterlogging disasters.Aiming at the lack of information mining ability of conventional precipitation forecasting method,a monthly precipitation combined forecasting model based on VMD-TCN was proposed.The model used variational mode decomposition(VMD)to decompose the original sequence into a series of relatively stable subsequence,and then uses temporal convolution network(TCN)to predict each subsequence respectively,output the prediction results of each subsequence,and obtained the final prediction result by superposition.The VMD-TCN monthly precipitation prediction model was compared with EMD-TCN model and VMD-LSTM model.The results show that the coefficient of determination(R^(2))of VMD-TCN model is 0.98;compared with EMD-TCN and VMD-LSTM,the root mean square error(RMSE)is reduced by 83.85%and 43.56%,respectively,the mean absolute error(MAE)is reduced by 84.25%and 43.60%,respectively,which is better than EMD-TCN and VMD-LSTM.On the basis of higher accuracy than VMD-LSTM model,the VMD-TCN model is based on the parallel thought of convolution,and its running speed is more than two times of VMD-LSTM model,which could provide an effective method for the prediction of monthly precipitation.