新疆玛纳斯河流域洪水预报研究

在充分考虑了融雪径流型河流自身特点的基础上,将自回归滑动平均模型ARMA(p,q),概化为ARMA(p,1),并运用最小信息(AIC)准则,确定模型的最佳阶数p,运用修正的可变遗忘因子递推最小二乘法,进行参数的动态修正,达到了最佳参数的跟踪效果.通过在玛纳斯河流域应用,表明该模型具有较高的预报精度,其确定性系数均达到了0.90以上,径流总量相对误差控制在7%以下,洪峰流量相对误差<10%,峰现时差≤2.

A Study of the Flood Forecast of the Manas River, Xinjiang

The Manas River, an internal river in arid district, rises in the high and cold mountains. There is a Kenswatt hydrometric station only. Because of lack of hydrological and meteorological data, it is very difficult to use determinacy model to do the hydrologic forecast. Therefore, it is necessary to set up a forecasting model with simple structure and fully utilize available hydrological and meteorological data. The model considers not only surface flow change but also temperature and rainfall factors that impact on surface flow randomly. The model is based on fully considering the characteristics of a snowmelt river, turning ARMA (%p, q%) generally into ARMA (%p%, 1) and using minimum information criteria (AIC) to determine the best steps (p) of the model. Applying the least square method of revised variable forgetting factor to identify the time-variation parameters and carry on the dynamic revision of parameters. Through applying to the Manas River, it is indicated that the model has a higher precision of forecasting. Its determinacy coefficient is greater than 0.90, its relative error of total runoff is less than 7%, and its relative error of discharge at flood peak is less than 10%. The time difference of peak appearance is less than 2.