关键水文信息对三峡水库调度决策的重要性分析

On the Importance of Key Hydrological Information on Dispatching Decision of Three Gorges Reservoir

城陵矶(莲花塘)水位为三峡水库调度决策的重要指标,采用随机森林算法建立三峡水库出库决策模拟模型,分析三峡水库历史运行数据,从多角度对三峡水库调度所采用的水文信息,如月份、水库水位、城陵矶水位、预报入库流量等,进行排名与判断其价值,并重点分析城陵矶水位对三峡水库调度决策的影响。研究表明,所采用的随机森林模型能够合理地模拟三峡水库的出库决策,汛期的模拟效果要好于非汛期,可能是汛期调度决策受到预报入库流量、城陵矶水位的影响更加直接;无论是全年、汛期还是非汛期,水文信息的重要性排名均为预报入库流量、城陵矶水位、水库水位、月份;预报入库流量在汛期的重要性高于非汛期。

The water level at Chenglingji(Lianhuatang) Station is an important index for the dispatching decision of Three Gorges Reservoir. A simulation model for the outflow decision of Three Gorges Reservoir is constructed with the random forest algorithm. The historical operation data of the reservoir is analyzed. Hydrological parameters used in the reservoir dispatching, such as the month, reservoir water level, water level at Chenglingji Station, and predicted inflow, are evaluated and ranked from multiple perspectives. Especially, the influence of the water level at Chenglingji Station on the outflow decision of the reservoir is analyzed. The results show that the random forest model can reasonably simulate the outflow decision of Three Gorges Reservoir. The simulation effect in flood season is better than that in non-flood season, possibly because the dispatching decision in flood season is more directly affected by the predicted inflow and the water level at Chenglingji Station. Whether for the whole year, flood season or non-flood season, the importance ranking of hydrological parameters is the predicted inflow, the water level at Chenglingji Station, the reservoir water level and the month. The importance of the predicted inflow in flood season is higher than that in non-flood season.