Study On The Application Of Xin’anjiang Model To Flood Forecasting For Small And Medium-sized Reservoir

Currently, it is difficult to measure direct runoff inflow into a reservoir via the shore, and it is usually lack of observed hydrologic data of main channels into the reservoir as well. However, the observed water level data is usually abundant for small and medium-sized reservoirs. Water level forecasting of small and medium-sized reservoir located in Auhui Province was chosen as research object, based on the DEM grid data extraction reservoir basin boundary, and the basin was divided into land surface and water surface, land surface basin is divided into several sub basins. Runoff yield on water surface could be calculated by multiplying precipitation and its area, with the data of water consumption and water discharge from a reservoir, then the inflow runoff to a reservoir yield on land surface could be back stepped. The flood forecasting model for small and medium-sized reservoir was established. Then genetic algorithm was used to calibrate the parameters of Xin’anjiang model, including day model and case model.Simulated results for Dongpu reservoir basin are:percent of pass of annual runoff depth was 92.9% for the calibration data of 14 years with the day model; percent of pass of annual runoff depth was 100% for the validation data of 18 years with the day model; percent of pass of runoff depth was 88.2% for 34 calibration floods, and percent of pass of runoff depth was 92.3% for 26 validation floods with the case model; percent of pass of water level forecasting of 60 floods was 90%. The accuracy of whole scheme reached to Class A. Simulated results for Luoji reservoir basin are:percent of pass of annual runoff depth was 75% for the calibration data of 8 years with the day model; percent of pass of annual runoff depth was 71.4% for the validation data of 7 years with the day model; percent of pass of runoff depth was 75% for 24 calibration floods, and percent of pass of runoff depth was 75% for 20 validation floods with the case model; percent of pass of water level forecasting of 44 floods was 79.5%. The accuracy of whole scheme reached to Class B. Simulated results for Shuanghe reservoir basin are:percent of pass of annual runoff depth was 70% for the calibration data of 10 years with the day model; percent of pass of annual runoff depth was 75% for the validation data of 8 years with the day model; percent of pass of runoff depth was 70% for 30 calibration floods, and percent of pass of runoff depth was 72% for 25 validation floods with the case model; percent of pass of water level forecasting of 55 floods was 70.9%. The accuracy of whole scheme reached to Class B.The calculation results show that the proposed flood forecasting model for small and medium-sized reservoir based on the usable observed hydrological data is feasible, and its accuracy can meet the requirement of reservoir operation. The proposed method can be applied to flood forecasting for similar reservoir basin.