基于1stOpt神经网络模型的年均径流预测与应用

针对年均径流预测问题，首先利用1stOpt软件进行逐个影响因子拟合，得到相关影响因子的主次顺序，提取关键因子在基于L-M算法和UGO算法下进行多元非线性曲线拟合；另外结合基于L-M算法的改进BP神经网络将所提取的关键因子进行径流预测，以北方某河流径流实例进行计算，两种方法效果均优于传统模型，且在多影响因子的情况下，组合模型精度和效果更好，为径流预测提供了一个新的更实用的方法。

Forecasting of annual runoff based on 1stOpt- neural network

Aiming to solve the problem of annual runoff prediction,the author fitted the factor one by one using the software 1stOpt,got the sequence of the related factors,extracted the key factors and fitted it with multivariate nonlinear curve fitting based on L-M algorithm and UGO algorithm;On the other hand ,the author predicted the annual runoff using the extracted key factors by the improved neural network based on L-M algorithm;With one of northern rivers runoff calculation example,it is proved the two methods were better than the traditional methods.And combination model is better in precision and effect when there are more related factors,providing a new and more practical method for annual runoff prediction.