Research on the Application of CEEMD-LSTM-LSSVM Coupled Model in Regional Precipitation Prediction

Abstract

Precipitation is a vital component of the regional water resource circulation system. Accurate and efficient precipitation prediction is especially important in the context of global warming, as it can help explore the regional precipitation pattern and promote comprehensive water resource utilization. However, due to the influence of many factors, the precipitation process exhibits significant stochasticity, uncertainty, and nonlinearity despite having some regularity. In this article, monthly precipitation in Zhoukou City is predicted using a complementary ensemble empirical modal decomposition (CEEMD) method combined with a long short-term memory neural network (LSTM) model and a least squares support vector machine (LSSVM) model. The results demonstrate that the CEEMD-LSTM-LSSVM model exhibits a root mean square error of 15.01 and a mean absolute error of 11.31 in predicting monthly precipitation in Zhoukou City. The model effectively overcomes the problems of modal confounding present in empirical modal decomposition (EMD), the existence of reconstruction errors in ensemble empirical modal decomposition (EEMD), and the lack of accuracy of a single LSTM model in predicting modal components with different frequencies obtained by EEMD decomposition. The model provides an effective approach for predicting future precipitation in the Zhoukou area and predicts monthly precipitation in the study area from 2023 to 2025. The study provides a reference for relevant departments to take effective measures against natural disasters and rationally plan urban water resources.