基于贝叶斯网络的流域内水文事件丰枯遭遇研究

Basin level study on synchronous-asynchronous encounter probability of hydrologic events based on Bayes-net theory

基于贝叶斯网络理论结合Copula函数建立了东江流域上、中、下游三个站点降雨、径流丰枯遭遇的风险管理模型,直观地描述了各个站点间降雨、径流的相互关系。利用Copula函数建立联合分布模型计算了站点间不同丰枯组合状态调水不利情况的风险概率。通过贝叶斯网络结构的反向推理功能,以后验知识作为输入,进一步对流域内未来调水可能面临的丰枯遭遇情况进行了仿真计算。结果表明,系统中一个节点的变化将会对其他节点的概率值产生巨大影响,以上、中游降雨为枯水情况作为后验输入,站点间丰枯遭遇调水不利风险概率增幅至55%以上;仿真结果可以为流域内调水方案的制定提供理论支撑。

A Synchronous-asynchronous encounter risk management model of rich-poor runoff and pr-ecipitation in the upper,middle and lower reaches of Dongjiang river basin was developed using a combined approach based on the copula function and Bayes-net works and the interrelation of runoff and precipitation among different three stations were intuitively described.Based on the Copula function,the joint distribution model was established to calculate the risk probability of the diversion adverse situation of Synchronous-asynchronous encounter.On the basis of a posterior knowledge input,the Backward reasoning function of Bayes-net was utilized to conduct simulation calculation of the potential Synchronous-asynchronous encounter states in some certain situations which may happen in the fututr.The result showrf that the change of a node probability value in the system would have a big impact on the others,on the premise that the rainfall shortage situation appeared in the upper and middle reaches,the risk probability of water transfer disadvantages,which increased to more than 55%;At last,the result of the simulating calculation would provide theoretical support to formulate the water transfer scheme in river basin.