中文摘要：

针对径流是典型非平稳序列这一特点及目前存在的径流中长期预测精度低的问题，提出一种新的耦合预测方法基于经验模态分解（Empirical Mode Decomposition ，EMD）与粒子群优化算法（Particle Swarm Optimization ， PSO）的非线性灰色Bernoulli耦合模型（Nash Nonlinear Grey Bernoulli Model ，Nash NGBM ）。首先利用EMD方法对汾河上游上静游、汾河水库、寨上和兰村4个水文站的年径流序列进行平稳化处理，分别得到若干个固有模态函数（In‐trinsic Mode Function ，IMF），对各阶IMF分别建立基于PSO算法的Nash NGBM（1，1）模型并进行预测，趋势项用多项式拟合并进行预测，然后通过重构各预测值得到汾河上游4个水文站年径流量的预测结果，并与单独运用基于PSO算法Nash NGBM （1，1）模型的预测结果进行比较，对模型作出评价。结果表明，基于EMD与PSO算法的Nash NGBM （1，1）耦合模型的拟合精度在92．5％以上，预测精度均达到了100％，其预测精度比单独运用基于 PSO算法 Nash NGBM （1，1）模型的预测精度有了明显提高；基于EMD与PSO算法的Nash NGBM （1，1）耦合预测方法的提出为径流中长期预测精度的提高提供了新的思路，对区域水资源的合理配置与优化调度具有重要的理论意义和实际价值。

英文摘要：

Considering the universally non-stationary property of runoff time series and the problem of low predicting accuracy of mid-long runoff forecasting ,a new time series prediction method-Nash NGBM （1 ,1） model based on Empirical Mode Decomposition （EMD） and Particle Swarm Optimization （PSO） ,is proposed in this paper .First ,the EMD is used to make the time series stationa‐ry ,obtaining the intrinsic mode function of different time scales .Then all the intrinsic mode functions is predicted by using Nash NGBM （1 ,1） model based on PSO .The residue by quadratic polynomial equation is predicted ,the predicted values are reconstructed as the predicted values of annual runoff .Lastly ,the predicting values are compared and evaluated with sole PSO-NGBM （1 ,1） model＆#39;s .The results show that：the fitting accuracy of Nash NGBM （1 ,1） model based on EMD and PSO is all above 92 .5% ,and the pre‐dicting accuracy is all up to 100% .Compared with the results of Nash NGBM（1 ,1） model based on PSO ,the fitting and predicting accuracy of this coupled model has been improved .The proposal of Nash NGBM （1 ,1） model based on EMD and PSO provides a new idea for the improvement of predicting accuracy of mid-long runoff forecasting .It is of practical significance to the rational allocation and optimal operation of water resources .