基于降雨径流相似性的短期径流预报研究

Research on Short-term Runoff Forecast Based on Rainfall and Runoff Similarity

径流预报是水力发电、防汛抗旱、水资源综合利用、电力生产与保障等的重要依据。针对物理驱动模型在参数率定困难和数据驱动模型预报结果可解释性差的问题,研究提出基于降雨径流相似性的短期径流预报模型,深入分析径流成因,采用数据挖掘技术剖析数据之间的关系以进一步提高预报精度;根据不同时期和量级的产汇流差异进行流量分级,并滚动更新降雨径流信息以实现滚动预报。该研究成果在乌江的应用显示,入库1日、7日预报结果平均相对误差分别为10.79%、14.39%,预报效果良好。

Runoff forecast is an important basis for hydropower generation,flood control and drought relief,comprehensive utilization of water resources,power production and security.In view of the difficulty in parameter calibration of physical-driven models and the poor interpretability of forecast results of data-driven models,a short-term runoff forecast model based on rainfall runoff similarity is proposed,which can analyze the causes of runoff in depth,and further improve the forecast accuracy by data mining technology to analyze the relationship between data.Through classifying the flow according to the differences of runoff yield and concentration in different periods and magnitude,and updating the rainfall runoff information in a rolling manner,the rolling runoff forecast is achieved.The application of the research results in Wujiang River shows that the average relative errors of the forecast results on the 1st and 7th day of the reservoir inflow are 10.79%and 14.39%respectively,indicating the forecast effect is good.