Study of regional monthly precipitation based on CEEMD-BILSTM coupled model

Abstract

The prediction of monthly precipitation is of great importance for regional water resources management and use. The monthly precipitation sequence is affected by various factors such as atmosphere, region and environment, and has obvious ambiguity, chance and uncertainty. CEEMD based on complementary ensemble empirical mode decomposition can effectively reduce the reconstruction error of time series, and bidirectional long short-term memory (BILSTM) model can effectively learn long-term dependencies in time series. A CEEMD-BILSTM (complementary integrated empirical mode decomposition-bidirectional long short-term memory) coupled model is constructed to predict the monthly precipitation in Zhengzhou, and the performances of the LSTM model, EEMD-LSTM model and EEMD-BILSTM model are compared. The CEEMD-BILSTM model has a maximum relative error of 7.28%, a minimum relative error of 0.00%, and an average relative error of 2.68%, with an RMS error of 2.6% and a coefficient of determination of 0.97 in predicting monthly precipitation in Zhengzhou, which is considered a good accuracy of the CEEMD-BILSTM model for predicting monthly precipitation in Zhengzhou. The model is better than the LSTM model, the EEMD-LSTM model, and the EEMD-BILSTM model and has better fitting ability. It also shows that it has strong nonlinear and complex process learning ability in the hydrological factor model of regional precipitation prediction. HIGHLIGHTS CEEMD is a novel data preprocessing method, which can effectively reduce the nonsmoothness of time series.; BILSTM models can efficiently learn long-term dependencies in time series.; CEEMD-BILSTM model has a higher prediction level, and the model is feasible for monthly precipitation prediction.;

Keywords

• bilstm neural network
• ceemd
• forecast accuracy
• precipitation forecast
• zhengzhou