Rainfall prediction in coastal hilly areas based on VMD–RSA–DNC

Abstract

Abstract Highly accurate rainfall prediction can provide a reliable scientific basis for human production and life. For the characteristics of occasional and sudden changes of rainfall in coastal hilly areas, this article chooses four cities in the eastern Zhejiang province as the object of the study and establishes a rainfall prediction model based on variational mode decomposition (VMD), reptile search algorithm (RSA), and differentiable neural computer (DNC). The VMD algorithm reduces the complexity of the sequence data; RSA is used to find the best-fit function; and DNC combines the advantages of the recurrent neural network and computational processing to improve the problem of memory forgetting of long short-term memory. To verify the prediction accuracy of the model, the prediction results are compared with the other three models, and the results show that the VMD–RSA–DNC model has the best prediction with the maximum and minimum relative errors of 9.62 and 0.17%, respectively, the average root-mean-square error of 5.43, the average mean absolute percentage error of 3.59%, and the average Nash–Sutcliffe efficiency of 0.95 for predicting four cities in the coastal hilly area. This study provides a new reference method for the construction of rainfall prediction models.