基于 MIV 的遗传神经网络径流预报模型

摘要

针对中长期水文预报中预报对象与预报因子之间复杂的非线性关系，引入平均影响值对预报因子进行筛选，选出对头道拐站年径流量影响较大的年降水量、年均相对湿度、年均气压3个因子作为神经网络的自变量，利用遗传算法优化的 BP 神经网络建立了预报模型。预报结果表明：基于平均影响值的遗传神经网络的预报精度及稳定性均达到了满意的效果。%In order to deal with the complex non-linear relationship between forecasters and predictors in mid-to-long-term hydrological forecasting, a mean impact value was introduced to screen forecasting factors,and the annual average precipitation,relative humidity and pressure were selected as the main factors to forecast the annual run-off in Toudaoguai sections with GA-BP neural network. The results show that both forecasting precision and stability can be satisfied with this GA-BP network based on MIV.