基于EEMD分解与BOA算法优化神经网络的密云水库大阁水文站径流预测

利用R/S分析法研究密云水库潮河流域大阁水文站1969-2013年的径流数据的变化趋势，以BP神经网络为背景，EEMD分解为辅助，建立分解-重构-预测的组合模型对月径流序列进行预测，利用蝴蝶算法（BOA）优化组合模型，综合得到最优预测模型。结果表明，大阁站年、月径流序列均呈现下降趋势；对月径流序列预测，BPNN预报合格率为60.0%，不能用于预报作业，但可作为参考使用（MAE=0.406，RMSE=0.539，MAPE=0.349 7）；引入BOA算法优化BP网络参数，得到EEMD-BOA-BP模型预报合格率为83.3%，可以用于预报作业（MAE=0.257，RMSE=0.347，MAPE=0.219 5）。通过EEMD分解得到分解-重构-预测组合模型对提高模型精度有一定的作用，同时在组合模型中引入优化算法能进一步提高模型精度。

Run-off Prediction of Dage Hydrological Station of Miyun Reservoir Based on EEMD Decomposition and Neutral Network Optimized by BOA Algorithm

The R/S analysis method was used to study the change trend of runoff data from 1969 to 2013 at the Dage Hydrological Station in the Chaohe Basin of Miyun Reservoir.With the BP neural network as the background,EEMD decomposition was assisted to establish a combination model of decomposition-reconstruction-prediction to predict the monthly runoff sequence.The butterfly optimization algorithm (BOA) was adopted to optimize the combination model to comprehensively obtain the optimal prediction model.The results of the study showed that the annual and monthly runoff series at the Dage Station showed a downward trend; for monthly runoff series prediction,the BPNN forecast pass rate was 60.0%,which could not be used for forecasting operations,but could be used as a reference (MAE=0.406,RMSE=0.539,MAPE=0.349 7); BOA algorithm was introduced to optimize the BP network parameters,and the EEMD-BOA-BP model forecast qualification rate was 83.3%,which met the requirement for forecasting operations (MAE=0.257,RMSE=0.347,MAPE=0.219 5).The decomposition-reconstruction-prediction combination model obtained through EEMD decomposition had a certain effect on improving the accuracy of the model.The introduction of the optimization algorithms into the combined model could further improve the accuracy of the model.