Runoff Prediction of Irrigated Paddy Areas in Southern China Based on EEMD-LSTM Model

To overcome the difficulty that existing hydrological models cannot accurately simulate processes with limited information in irrigated paddy areas southern China, this paper presents a prediction model combining Ensemble Empirical Mode Decomposition (EEMD) method and Long Short-Term Memory (LSTM) network. Meteorological factors were set as multivariate input to model. Rainfall, regarded main variable affecting runoff, was decomposed reconstructed into combination of new series stronger regularity by using EEMD K-means algorithm. The LSTM used explore data laws then predict runoff areas. Yangshudang (YSD) watershed Zhanghe Irrigation System (ZIS) Hubei Province, China taken study area. Compared other models, results show EEMD-LSTM had better simulation performance, an NSE above 0.85. Among them, R2, NSE, RMSE RAE EEMD-LSTM(3) best, they 0.85, 0.86, 1.106 0.35, respectively. accuracy peak flows than well performance rainfall nonrainfall events, while improving 0.05, 0.24 0.24, respectively, compared EEMD-LSTM(1) Overall, multivariations is suited for simulating predicting daily-scale rainfall–runoff process China. It can provide technical support help decision making efficient utilization management water resources.