基于EWT-PSO-Elman耦合模型在径流预测中的应用

由于径流序列的非线性和非平稳性,单一预测模型能力有限,难以做出准确预测。因此,论文基于澄碧河流域坝首站1979-2019年共41a的实测月径流序列,引入经验小波变换分解(EWT)、粒子群算法(PSO),建立一种基于Elman神经网络的组合月径流预测模型(EWT-PSO-Elman),并采用纳什效率系数(NSE)、平均相对误差绝对值(MAPE)和均方根误差(RMSE)对测试集的预测结果进行评价与分析,并将预测结果与EWT-PSO-BP、PSO-Elman、PSO-BP、Elman、BP模型进行比较。结果表明：EWT-PSO-Elman模型的纳什效率系数为0.9135,均方根误差为19.4511,预报等级为甲级,具有较好的预测精度和泛化能力；EWT-PSO-Elman模型的预测精度优于EWT-PSO-BP、PSO-Elman、PSO-BP、Elman、BP模型。可见,EWT-PSO-Elman模型具有更好的预测精度,可应用于径流预测研究中。

Application of EWT-PSO-Elman based coupled model in runoff prediction

Due to the non-linear and non-stationary nature of the runoff series, a single prediction model has limited ability to make accurate predictions. Therefore, the paper is based on a total of 41a of measured monthly runoff series from 1979-2019 at the dam head station in the Chengbi River basin, An empirical wavelet transform decomposition (EWT) and particle swarm algorithm (PSO) were introduced to establish a combined monthly runoff prediction model (EWT-PSO-Elman) based on Elman neural network, The Nash efficiency coefficient (NSE), mean relative error absolute (MAPE) and root mean square error (RMSE) were also used to evaluate and analyze the prediction results of the test set, and the prediction results were compared with the EWT-PSO-BP, PSO-Elman, PSO-BP, Elman, and BP models. The results show that: The EWT-PSO-Elman model has a Nash efficiency coefficient of 0.9135, root mean square error of 19.4511 and a forecast grade of A, which has good prediction accuracy and generalization ability; The prediction accuracy of the EWT-PSO-Elman model is better than that of the EWT-PSO-BP, PSO-Elman, PSO-BP, Elman, and BP models. It can be seen that the EWT-PSO-Elman model has better prediction accuracy and can be applied to runoff prediction studies.