人工神经网络在水文资料插补延长中的应用

水文资料的插补延长一直是水文计算中的一个难题.本文针对水文资料的插补和水文资料的延长问题进行系统的研究.插补水文资料时,采用人工神经网络双向时间序列插补模型,打破了传统的单向时间序列识别模式,应用缺测时段前后已知时段水文资料,插补出缺测水文资料;展延长系列水文资料则应用人工神经网络参证站模型,并应用流量较大年份的径流资料预测未知年份的径流资料,来进一步提高预测精度,并结合紫坪铺流量资料插补延长实例,检验模型的可行性.结果表明该模型对水文资料的插补或对未知年份的径流量都能够进行较好的预测.

Application of Artificial Neural Networks to Interpolation and Extrapolation of Hydrological Data

The problem in estimating missing hydrological data is always paid an attention. In this paper, two ANNS models, the double - directions interpolating model and the referenced station model are proposed and explained. The double - direction interpolating model of ANNS, different from the traditional single - direction model, is applied to estimate the missing data with available value before and after the gap. The referenced station model of ANNS is applied to extend records by adding lengthy segments of estimated data. In addition, how to improve forecasting precision and the applied feasibility, the bigger data series is used for training while the smaller data series is used for checking. In this paper, stream flow at Zipingpu station is tested daily. The accuracy of the result is superior to that obtained by iterative regression analysis. The test results prove that ANNs can be applied successfully to interpolate and extend the hydrological data.