根据1960 年- 2011 年的实测入库流量资料, 以河川径流量为相依随机变量, 介绍加权马尔可夫链模型的相 关概念及预测未来一年入库流量的步骤, 采用均值2标准差分级法把入库流量序列划分成枯、偏枯、偏丰、丰4 种状 态。以各阶自相关系数为权重, 预测2010 年- 2011 年的入库流量, 将其所在状态区间与实测值进行对比。结果表 明, 加权马尔可夫链模型对密云水库入库流量预测精度较高, 以此又对2012 年- 2013 年的入库流量进行了预测。 最后对其遍历性和平稳分布进行分析, 计算入流丰、枯状态在实测序列中的重现期, 其中出现偏枯状态的概率最大, 由此预测密云水库未来的入库流量处于偏枯状态。

Acco rding t o the act ual inflow data of the Miyun Reservo ir fr om 1960 to 2011, riv er runoff was select ed as the r andom variable, and t he r elated concept of the w eig hted Markov chain model and the st eps for the inflow pr ediction in the incoming one year w ere intro duced. The classificat ion metho d of averag e2standard was used to div ide the inflow sequence into four conditio ns, including dr ought, lean dr ought, lean wet, and w et. The autocor relation w as r egar ded as weig ht coefficient to pr edict inflow be2 tw een 2010 and 2011, which w ere compared w ith the measured data. T he results showed that the weig hted Mar ko v cha in model can predict inflow o f the Miyun Reserv oir w it h high pr ecision. Therefo re, the model w as used to predict inf low betw een 2012 and 2013. Finally, the erg odicit y and statio nary distr ibut ion of Markov cha in wer e analyzed, and the return per iods of o bser ved sequence under the w et and dry condit ions w ere calculat ed, which sugg ested that the occurr ence pro bability o f lean dro ught is the larg est. The inflow of the Miy un Reserv oir was predicted to be lean dr ought in the futur e.

北京市科技计划课题( Z141100006014049) ; 国家科技重大专项课题( 2012ZX072052005)