Hybrid short-term runoff prediction model based on optimal variational mode decomposition, improved Harris hawks algorithm and long short-term memory network

Abstract

Abstract Runoff prediction is an important basis for rational allocation of basin water resources and plays a very important role in regional water resources management. In this study, a hybrid short-term runoff prediction model based on long short-term memory network (LSTM), improved Harris hawks optimization algorithm (IHHO) and optimal variational mode decomposition (OVMD) are proposed. Firstly, the original runoff data is decomposed into several sub-modes by OVMD, and then the sub-modes are reconstructed by phase space reconstruction (PSR). Secondly, the Harris hawks optimization algorithm is improved by the chaos map and the hill climbing algorithm. Then, the LSTM model is established for each sub-mode, and the improved Harris hawks optimization algorithm (IHHO) is used to optimize the number of hidden layer neurons and learning rate of the LSTM network. Finally, the results of all sub-modes are combined to obtain the finally runoff prediction result. In this study, seven control models are constructed and compared with the proposed model to verify the effectiveness of the proposed model in runoff prediction.