论文部分内容阅读
小波回归模型(WR,Wavelet Regressive model)是目前常用的且性能较优的一种水文时间序列模拟预报方法.然而,小波回归模型不能很好地对水文序列预报结果进行不确定性分析和评估,给水文决策等工作带来较大的风险.开展水文概率预报并定量估计预报结果的不确定性是更加合理的做法,它可以将不确定性和风险很好地纳入决策和政策制定中.本文将自适应性马尔可夫链蒙特卡罗采样方法(AM-MCMC,Adaptive Metropolis-Markov chain Monte Carlo)应用到小波回归建模过程中,提出了一个水文时间序列概率预报的新模型,称为AM-MCMC-WR模型.在该模型中,AM-MCMC采样方法主要用于估计和定量描述小波回归模型中参数的不确定性,基于此可以实现水文时间序列的概率预报.利用AM-MCMCWR模型进行预报时主要分4个步骤:序列离散小波分解、确定小波回归模型、参数AM-MCMC采样及不确定性
Wavelet Regressive model (WR) is a commonly used and better performance hydrological time series simulation and forecasting method, however, the wavelet regression model can not well analyze and evaluate the uncertainty of hydrological sequence forecasting results, Bring greater risks to hydrological decision-making, etc. It is a more reasonable approach to carry out hydrological probability forecast and quantitatively estimate the uncertainty of forecast results, which can well incorporate the uncertainty and risk into decision-making and policy-making. In the process of wavelet regression modeling, AM-MCMC (Adaptive Metropolis-Markov chain Monte Carlo) is used to propose a new model of hydrological time series probabilistic prediction, which is called AM -MCMC-WR model.In this model, the AM-MCMC sampling method is mainly used to estimate and quantitatively describe the uncertainty of the parameters in the wavelet regression model, based on which the probability prediction of hydrological time series can be realized.Using the AM-MCMCWR model Prediction is mainly divided into four steps: discrete wavelet decomposition sequence to determine the wavelet regression model, the parameters of the AM-MCMC sampling and uncertainty