Inflow combination forecast of reservoir based on SWAT, XAJ model and Bayes model averaging method

Abstract

Abstract Inflow forecast plays an indispensable role in reservoir operation. Accuracy and effectiveness of model prediction play a decisive role in it. In this paper, the certainty coefficient, root mean square error (RMSE), mean absolute deviation (MAE) and Nash-Suthcliffe coefficient (NSE) are used to consider the effect of Soil and Water Assessment Tool (SWAT) and Xin'anjiang (XAJ) model on the inflow prediction of Jinxi Reservoir. Results indicate that the certainty coefficient of XAJ model is 90.95% and SWAT model is 84.98%. The NSE of SWAT model is 84.42% and the NSE of XAJ model is 82.33% in which the NSE of the SWAT model is superior to the XAJ model. In view of this, based on the two hydrological models results, Bayes Model Averaging (BMA) method is used to study the inflow combination forecast. The SWAT-XAJ coupling model based on BMA method is proposed. During the calibration period and the verification period, compared with single model, the maximum increment of the deterministic coefficient of BMA is 2.87%, the maximum decrement of RMSE is 82.00, and the maximum increment of NSE is 6.80%. This shows that the new model can effectively overcome the shortcomings of single hydrological model and give full play to the advantages of different forecasting models.