基于EEMD与ANN混合方法的水库月径流预测

Monthly Runoff Forecasting of Reservoir Based on EEMD and ANN Hybrid Model

为了解决径流序列复杂的非稳态特征并提高径流的预报精度,采用EEMD-ANN组合方法构建径流预报模型,其中EEMD方法通过将非线性非稳态的水文序列分解为多组固有模态分量及趋势项,实现径流序列的稳态化,然后使用ANN方法分别进行预测,进而完成径流序列重构。以黄河龙羊峡水库为例,基于EEMD-ANN预报模型对入库径流量进行了预测,结果表明该方法可较精准地预测径流量。同时,通过对比分析发现,采用EEMD-ANN连续滚动预测月径流量在汛期的预报效果较好,而非汛期可采用同期预报的手段提高径流预报精度。

In order to improve the prediction accuracy, ensemble empirical mode decomposition(EEMD) and artificial neural network(ANN) hybrid model was proposed for monthly runoff forecasting, which considered runoff series complicated non-stationary characteristics. Firstly, in order to achieve steady state of runoff series, the non-linear and non-stationary runoff series was decomposed into several intrinsic mode functions(IMFs) and a trend series by using EEMD. Then, ANN model was established for different IMFs and the trend series respectively. Lastly, reconstructing forecast runoff sequence by superimposing all forecasting model. Taking Longyangxia(LYX) Reservoir of the Yellow River as an example, the inflow was predicted based on EEMD-ANN model. The results indicated that the hybrid method could accurately forecast the inflow of LYX Reservoir. At the same time, through comparison continuous and adaptive forecast and runoff series of same month forecast based on EEMD-ANN, it was found that the former method was better in flood season for forecasting monthly runoff, while runoff series of same month forecast method could be used to improve the accuracy of runoff prediction in dry season.