基于AM-MCMC算法的贝叶斯概率洪水预报模型

本文在贝叶斯预报系统的框架下,利用BP网络能描述非线性映射的特性建立了基于BP网络的先验密度和似然函数的模型,并采用基于自适应采样算法(Adaptive Metropolis algorithm,简称AM)的马尔可夫链蒙特卡罗模拟方法(Markov Chain Monte Carlo,简称MCMC)求解流量的后验密度,最后给出流量的概率预报。实例表明,基于AM-MCMC的BP贝叶斯概率水文预报的精度高,且能给出预报的方差,使得防洪决策可以考虑预报的不确定性。

Bayesian probabilistic flood forecasting model based onadaptive metropolis_MCMC algorithm

The models of prior density and likelihood function based on BP algorithm for ANN were built under the general frame of Bayesian forecasting system(BFS).The posterior density of discharge was obtained from Markov Chain Monte Carlo(MCMC) simulation method based on the adaptive metropolis(AM) algorithm.The probabilistic forecasting was applied to a case study.The result shows that the accuracy of forecasted discharge is higher than that of the Xinganjiang model,and the variance of prediction can be given at the same time.By using this method the uncertainty of prediction can be considered in the decision-making for flood prevention.