基于小波变换和GRNN神经网络的径流模型在雅砻江流域中的应用

Application of Runoff Model on Wavelet Conversion and GRNN in Yalongjiang River

对雅砻江甘孜站控制流域月降水量和气温做 Harr小波变换 ,并作为 GRNN神经网络的输入 ,对雅砻江甘孜站径流进行模拟和预测验证 ,效果较好。但该模型对气温变化不敏感。故应用不引进小波变换的 GRNN神经网络模型 ,并采用全球变化成果 ,在不同的气候情景下 ,对甘孜站径流进行预测。结果表明 ,在未来一段时间内 ,甘孜站径流量会有一定程度的增加。径流量对气温的响应不同于西北内陆河流域 ,原因是该流域月径流量随气温的升高而增加。可利用该模拟对甘孜站径流进行反延或预测 ,并利用甘孜、温波二站径流量关系 ,计算温波站径流量 。

Yalongjiang river is one of the three rivers that are included in the West-line Programme of Pumping Water from the South to the North of China, but there are only five-year runoff datum at Wenbo station, where will build a pumping dam. The Ganzi Hydrologic station,locating in the lower part of Yalongjiang River,has collected relatively long-term datum. Thus we can use the correlation between the runoff at Ganzi station and that of the Wenbo station,to calculate the runoff at Wenbo. Therefore, we should firstly extrapolate the runoff series at Ganzi. Since wavelet conversion and neural network are used extensively recently, we use Haar wavelet and GRNN neural network to simulated and predicted the runoff at Ganzi station of Yalongjiang River. The basic function of Haar wavelet is: Ψ(x)=1 0≤x≤0.5 -1 0.5≤x≤1 (1) 0 其他 The scaling function is: phi(x)=1 0≤x≤1 0 others (2) The reconstruction function of Hydrologic series r(t) is: r(t)=D 3(t)+∑3i=1A i(t)(3) Where D is reconstructed approximation coefficent, and CA is reconstructed detail coeffcient. See the algorithm of GRNN neural nrtwork in reference . We have monthly runoff converted, acquire the reconstructed approximation and detail coefficients,and use them as input samples of the GRNN neural network. The results are perfect. However,this method can't be used to predict the runoff in the future. For this reason, we take the second path. Let the precipitation and air temperature converted, then use them as the input samples of GRNN neural network, and the runoff as output samples. The simulated and predicted results are all perfect. Now we use the global changing results to predict the changes of yearly runoff, but the monthly runoff and yearly runoff are all obtuse to the monthly air temperature. There fore, we use GRNN without wavelet conversion. Now are the results. If precipitation is unchanged, the air temperature arises 0.5℃ and 1℃ respectively, the yearly runoff at Ganzi station of Yalongjiang River may decrease 6.35% and arise 8.74%, If air temperature does not change and precipitation arises 10% and 20%, the yearly runoff will arise 4.00% and 26.61 respectively. If air temperature airses 0.5℃, and precipitation arises 10% at the same time, the yearly runoff would arise l3.14%. From these results, we can conclude that the yearly runoff at Ganzi station of Yalongjiang River should arise in the near future. The response of runoff at Ganzi station of Yalongjiang River to air temperature disagrees with that of inland basins of Northwest China. The reason is that,in Yalongjiang River basin, when the air temperature arises, the runoff arises. At last, we can conclude that the model be suitable to the Yalongjiang River basin. Therefore, we can use the calculated results of Ganzi station to calculate the runoff series at Wenbo station of Yalongjiang River.