基于贝叶斯概率预报的水库汛限水位实时动态控制研究

Real-Time Dynamic Control of Reservoir Flood Limited Water Level Based on the Bayesian Probabilistic Forecasting Processor

摘要: 实时洪水预报精度较高是实施汛限水位动态控制的基础，其不确定性又是动态控制的主要风险源。针对洪水预报不确定，采用贝叶斯水文不确定性处理器将确定性预报转化为概率预报；针对入库洪水预报精度较高，建立了考虑未来洪水大小的三峡水库汛限水位实时控制模型。应用结果表明，相对于原设计调度方案，实时动态模型根据预报洪水大小执行不同的预蓄预泄策略，在不增加防洪风险的前提下可显著地提高水库的发电效益。基于贝叶斯修正的概率预报能够为决策者提供更多的洪水预报信息，有利于调度人员实施汛限水位实时动态控制运用方式。

Abstract: A real-time dynamic control of reservoir Flood Limited Water Level (FLWL) model was proposed based on the Bayesian forecasting processor. In terms of relative high inflow forecast precisions of the Three Gorges Reservoir (TGR), the different pre-release and refill strategies were implemented according to the reservoir inflow forecasting results. On account of inflow uncertainty, the deterministic inflow forecasts were transformed to probabilistic forecasts. Case study results indicate that the proposed model can greatly increase hydropower generation compared with current operation rules. It is also shown that the proposed model could make use of the future inflow magnitudes and provide more information for decision-makers, which could be in favor real-time and dynamic control of reservoir FLWL.