中文摘要：

目前河川径流、水库入库径流等情势预测广泛应用马尔柯夫链方法，并取得了很多研究成果，但多数研究侧重以各阶自相关系数为权重，进行加权预测，以期改进预测结果，对影响结果的其他原因缺乏探究．从马尔柯夫链方法中的变量的状态划分着手，分别采用了均值均方差法、保证率法、百分比法3种状态划分方法，探讨了不同状态划分方法对预测方法和预测结果的影响，建立了湖北高关水库年入库径流量的马尔柯夫链情势预测模型，研究表明常用的均值均方差径流状态划分方法并不适用于高关水库．建立的马尔柯夫链模型预测结果表明，预测效果较好．

英文摘要：

Many studies had been done on the application of Markov chain to prediction of state of annual runoff or precipitation. To improve the result of prediction, a weighted Markov chain prediction model is established by using several order of autocorrelation coefficient in many studies. Influences of three meth- ods of state division on establishing model and results of prediction are studied. It is proved that the mean and standard deviation method can＇t be used for Gaoguan Reservoir. Established Markov chain model is used to forecast the annual inflow state of Gaoguan reservoir; the results indicate that this method has good precision.