基于ε-SVR的草灌被覆坡面流阻力模型研究

Study on flow resistance coefficient model of overland covered by grass and shrub based on ε-SVR

坡面流阻力是反映坡面流水力学特性的重要特征值。通过人工模拟降雨试验,获取45 mm/h、90 mm/h和130 mm/h降雨强度下,20°坡面草地和灌木地的坡面流水力学参数。以模拟降雨雨强、坡面流流速、坡面流水深、坡面糙率为输入,坡面流阻力系数为输出,分别建立了草地、灌木地坡面流阻力系数ε-SVR模型,使用网格搜索和9折交叉验证进行模型参数寻优。结果表明,草地、灌木地坡面流阻力ε-SVR模型的MSE分别为0.028、0.05,模型对训练集的平均拟合误差分别为4.68%、5.39%,对测试集的平均预测误差分别为7.35%、4.65%,模型总体预测精度较高,具有较好的逼近能力和泛化能力。该研究对于完善草地、灌木地坡面流阻力计算方法,提高侵蚀预报模型的预报精度等具有一定意义。

Overland flow resistances on grass slope and shrub slope were studied on a 20° slope by rainfall experiment. The model of overland flow resistance coefficient is established based on ε-SVR, model parameters optimized by 9-fold cross-validation and grid search algorithm. The results showed that the model tend to be higher simulation accuracy and approximation and generalization ability. The model MSE of grassland and shrub land were respectively 0. 028 and 0.05, the average prediction error of training sets were respectively 4.68% and 5.39% , the average prediction error of test sets were respectively 7.35 % and 4.65%. The objective of this study is essential to improve the calculation methods of overland flow resistance coefficient and the prediction accuracy of erosion model.