基于BP神经网络的场次洪水河道沙量预报

Flood and sediment prediction based on BP neural network

针对日益严重的河流泥沙问题,掌握河流泥沙的影响因素和泥沙的变化过程是泥沙治理的关键。以神经网络模型为基础,建立场次洪水沙量预报模型,对多沙河流的洪水挟沙量进行预报,并取得较好的预报效果。选择辽西北多沙河流大凌河作为研究实例,首先将1984年-1998年间的29场历史实测洪水资料进行分析,得到影响下游沙量的主要因素;然后,通过神经网络模型建立上游影响因素与下游沙量之间的关系;最后,选取其中6场洪水资料进行验证。模型计算结果表明,计算结果与实测结果误差在合理范围之内,精度符合要求,可以用于下游沙量的预报。

Knowledge of the impact factors and variation process of river sediments is the key to solve the increasingly serious river sediment problems. In this paper, flood and sediment prediction model was developed to forecast the sediment load based on artificial neural network, which generated promising results. The model was then applied to the Daling River in the northwest of Liaoning Province. First, the data from 29 historical flood events from 1984 to 1998 were analyzed using the statistical method to obtain the main impact factors of downstream sediment load. Then, the BP neural network model was developed to characterize the relationship between the upstream impact factors and downstream sediment load. Finally, the data from six flood events were used to verify the model. The results showed that the errors between the calculated and measured values are within the reasonable range and meet the accuracy requirement, therefore the model is applicable for downstream sediment prediction in the Daling River.