基于KG-BP神经网络在秦淮河洪水水位预测中的应用

为了提高BP神经网络模型的预测精度,提出了一种基于KNN算法及GA算法优化的BP神经网络的水位预测方法(KG-BP),即通过KNN邻近算法从全样本数据中剔除与待测点相关度较低的样本集,并允许保留K个"优质"训练数据集;将筛选出的"优质"训练数据集代入GA算法中实现初始权阈值的优化;再将"优质"的样本和初始权阈值代入BP模型中进行训练。将该预测方法应用于东山站水位实际预测中,并与BP模型、GA-BP模型的预测结果进行对比分析,验证了KG-BP模型具有较高的预测精度。

Application of KG BP Neural Network in Flood Forecasting of Qinhuai River

In order to improve the prediction accuracy of BP neural network model, this paper proposed a water level prediction method based on KNN algorithm and GA optimized BP neural network. The KNN proximity algorithm removed the sample set with lower correlation between the sample data and the point to be measured, and retained the number of k high quality training data sets. The selected high quality training data set was substituted into the GA algorithm to optimize the initial weight threshold. And then the high quality sample and initial weight threshold were used to train the BP model. The prediction method was applied to the actual water level prediction of Dongshan Station, and compared with the prediction of BP model and GA BP model. It was verified that the KG BP model has a high prediction accuracy.