基于四种方法的月径流预测研究

径流预测对于水资源的合理开发利用与统筹配置具有重要意义。根据黄土高原地区渭河支流-北洛河状头水文站和泾河张家山站的月径流资料,运用门限自回归模型、神经网络模型、方差分析外推法以及季节水平模型四种方法对其进行预测,观察模拟效果并比较各自优缺点。对于枯水期月径流,季节水平模型对于两站预测合格率均为100%;方差分析外推法对于状头站和张家山站预测合格率分别为90%,80%;门限自回归模型对于两站的预测合格率均为80%;神经网络模型预测两站汛期月径流合格率均为100%。表明季节水平模型适用于枯季月径流的预测,神经网络模型适宜于汛期月径流预测,并且精度良好。

Study on Monthly Runoff Prediction Based on Four Methods

Runoff forecast is very important to the rational utilization and distribution of water resources.According to the monthly runoff data from Zhuangtou Hydrologic Station of Beiluo River and Zhangjiashan Hydrologic Station of Jinghe River,which are two branches of Weihe River in loess plateau,the threshold auto-regressive model,neural network model,variance analysis extrapolation as well as the seasonal level model are used to predict the monthly runoff,observe the similation results and find their advantages and disadvantages.The results show that in a dry season,the eligible rates for runoff forecast by using the seasonal level model for the two stations are both 100%.Using the variance analysis extrapolation for Zhuangtou Station and Zhangjiashan Station,the eligible rates are 90% and 80% respectively.The eligible rates for runoff forecast by using the threshold auto-regressive model for the two stations are both 80%.While in a flood season,the eligible rates for runoff forecast by using the neural network model for the two stations are both 100%.This study show that the seasonal level model is applicable to the runoff forecast in a dry season.And the neural network model is suitable for the runoff forecast in a flood season,and both models have a good simulation accuracy.