摘要
建立线性自回归模型,应用于洪水实时预报,并应用AIC、BIC这两种准则以确定自回归模型的阶数。最小二乘递推算法是估计自回归参数的一种常见方法。最小二乘法估算出的模型参数在预报误差平方和最小的条件下是最优解。研究中,为了强化时变系统的辩识以提高洪水预报精度,对数据采取衰减记忆、有限记忆及时变衰减记忆的方式,对基本的最小二乘递推算法提出了三种改进形式,并利用这几种改进算法进行了洪水演算,最后对几种算法的演算结果进行了比较。
A linear AR model is set up to be applied in real-time flood forecasting. Two criterion AIC and BIC are used to decide the exponent number of the linear AR model. We often use RLS procedure to estimate the parameters of AP model. The estimated parameters using RLS procedure are the optimum solution based on the condition that the sum of square of the forecasted errors is the minimum. In research, three improved RLS procedures are developed to strengthen the characteristic-identification of the time-varying system in order to increase the accuracy of flood forecasting. The three procedures are operated separately. And the calculated results using the three procedures are compared.
出处
《水力发电》
北大核心
2006年第8期14-16,共3页
Water Power
基金
国家自然科学基金资助项目(50479017)
关键词
线性自回归模型
最小二乘递推算法
有限记忆
自适应衰减因子
实时洪水预报
exponent number of linear AR model
faded-memory RLS procedure
fixed-memory RLS procedure
adaptive faded-memory
factor RLS procedure
real-time flood forecasting
