考虑水文模型参数时变的水库水位多预见期实时预报——以水布垭水库为例

Real-time forecasting of reservoir water levels over multiple lead timesconsidering time-varying hydrological model parameters andits application in the Shuibuya Reservoir

精准可靠的坝前水位预测可支撑水库的调度决策,保障流域防洪安全。传统水库水位实时预报方法均基于降雨径流稳态假定,在变化环境下精度不高。为此,本研究提出了一种整合水文模型参数时变和实时校正的水库水位多预见期实时预报方法:识别变化环境下水文模型参数的时变过程,构建时变参数与自然、社会经济因子的函数关系;基于时变参数函数式构建预报调度集成模型,以预见期内多个时刻预报水位与观测水位的吻合程度为目标函数率定水文模型的参数;采用基于反向拟合算法的实时校正技术进行误差校正,输出不同预见期的水库水位预报。以水布垭水库流域为研究对象,结果表明,较传统水库水位预报方法,论文提出的方法在率定期和检验期的平均绝对值误差分别下降了0.050和0.040 m,可为防洪减灾提供高精度、长预见期的水库水位预测。

Accurate and reliable reservoir water level forecasting supports reservoir decisions and ensures the safety of flood control in watersheds.Conventional methods for real-time forecasting of reservoir water levels are based on the stationary assumption of the rainfall-runoff relationship,the performance of which decreases in the changing environment.To this end,a real-time reservoir water level forecasting method over multiple lead times is developed,which integrates a hydrological model with time-varying parameters and error corrections:the time-varying parameters,and functions are constructed between the time-varying parameters and natural socio-economic factors;the time-varying parameter functions are embedded in the integrated hydrological and reservoir routing model,and the model is calibrated by minimizing the sum of the squared errors between forecasted and observed streamflow over multiple lead times;and an error correction method based on a back-fitting algorithm is used to update the forecasting of the reservoir water levels over multiple lead times.The results of the case study in the Shuibuya reservoir basin show that the proposed method compared with the conventional method has average absolute errors decreasing by 0.050 m and 0.040 m during the calibration period and the validation period,respectively,which can provide high accurate reservoir water level forecasting over long lead times for flood control.