温州市年降水量的权马尔可夫链预测模型

An Annual precipitation Forecasting Based on Weighted Markov Chain in Wenzhou

基于降水过程不确定性的特点,利用资料系列的均方差方法把降水序列分为不同的状态。在此基础上,采用统计方法,建立转移概率矩阵,以规范化的各阶自相关系数为权重,并用加权的马尔可夫模型和模糊集理论中的级别特征值预测降水状态和降水量。根据温州市区1955—2004年连续50 a的降水资料,用马尔可夫预测模型预测温州市2005—2007的降水状态及年降水量,预测误差分别为1.44%、0.3%、6.3%,满足预报精度,说明该模型应用于温州市区是成功的。

Based on the uncertainty characteristics, rainfall series could be divided into different states via mean square deviation method of data series. Transition probability matrix is obtained by using statistical method on that ＇basis. Standardized autocorrelafion e^ffieients based on the special characteristics of correlation among the historical stochastic variables are regarded as weights. The weighted Markov chain model and file level characteristics value of fuzzy sets are used to predict the probability state and the precipitation. Furthermore, according to 50 - year continuous precipitation data from 1955 to 2004, the improved model is used to predict annual precipitation from 2005 to 2007. The forecasting errors are 1.44%, 0.3% and 6.3%, respectively, which meets the request of predication precision. It shows the model is successfully applied in Wenzhou.