基于流水线模式的河系径流预报并发计算研究

Research on concurrent computing of river runoff forecast based on pipeline model

在河系径流预报计算中,受单站水文过程计算复杂性及下游站点依赖上游关联节点的制约,现有洪水预报系统多采用串联模式计算。但在河系预报节点多、模型方法复杂时,串联模式计算效率较低。为提高河系径流预报计算效率,引入流水线并行模式,将河系径流预报站点初始化、单元产汇流计算、河道洪水演算、校正分析等模块进行拆解,构建流水线式工作站,将径流预报站点按水力联系连续入站,实现河系节点集径流过程的平行并发计算。选取淮河正阳关以上流域50余个断面进行模拟试验,结果表明:构建的并发计算方法结果可靠,与串行模式结构相比,并发计算效率提升3倍多,能够有效提高河系径流预报效率,满足洪水预报实时性要求,尤其适用于B/S模式对系统响应效率的需求。

Affected by the complexity of single-station hydrological process calculation and the dependence of downstream stations on upstream related ones,the series mode is mostly applied in existing flood forecast systems.However,the series model is low-efficient in the case of too many prediction nodes and complicated model method.In order to improve the calculation efficiency,this paper introduces the idea of concurrent pipeline model,and disassembles the modules of forecast station initialization,unit runoff calculation,river flood routing and correction analysis.A pipeline workstation is built,where the forecast stations are input continuously according to hydraulic connection,aiming at realizing parallel concurrent calculation of the node runoff.A study case is carried out in the watersheds above Zhengyangguan of the Huaihe River,where there are more than 50 runoff sections.The case results show that the proposed concurrent calculation method is reliable,and the efficiency of concurrent calculation is more than 3 times higher than that of the serial model.Thus,it can effectively improve the efficiency of river runoff forecast,and meet the real-time requirements,especially the B/S model's demand for system response eficiency.