基于VMD-SSA-LSTM的月径流预测模型及应用

Monthly Runoff Prediction Model Based on VMD-SSA-LSTM

为提高月径流预测精度,提出了变分模态分解(VMD)和麻雀搜索算法(SSA)与长短期记忆神经网络(LSTM)相耦合,建立了月径流预测模型(VMD-SSA-LSTM)。首先利用VMD对历史径流数据进行分解,然后依据SSA对LSTM的参数进行寻优,并将分解出的月径流分量输入到LSTM神经网络,最后将每个分量的预测值相加,得到月径流预测值,并以福建池潭水库1950~2019年的月径流数据对模型进行验证。结果表明,与LSTM、VMD-LSTM模型相比,VMD-SSA-LSTM模型的预测精度更高,为开展月径流预测工作提供了一种新的选择。

In order to improve the accuracy of monthly runoff prediction, the variational mode decomposition(VMD) and sparrow search algorithm(SSA) were coupled with the long short-term memory neural network(LSTM) to establish a monthly runoff prediction model(VMD-SSA-LSTM). Firstly, the historical runoff data were decomposed through VMD method. Then the SSA algorithm was adopted to optimize the parameters of the LSTM, and the decomposed results were inputted to the LSTM neural network. Finally, the monthly runoff prediction value can be obtained by superposing the predicted value for each component. The model was validated with the monthly runoff data of Chitan Reservoir in Fujian from 1950 to 2019. Compared with the LSTM and the VMD-LSTM models, the results show that the proposed VMD-SSA-LSTM model has a higher prediction accuracy, which provides an alternative method for the monthly runoff prediction.