人工神经网络模型在黄河水质预测中的应用

Prediction of the Yellow River Water Quality Based on Artificial Neural Network

采用画匠营子断面2004—2009年逐周水质指标资料作为神经网络模型的训练样本,对BP神经网络进行训练,分别建立了pH值、溶解氧、氨氮、高锰酸盐指数的预测模型。为了验证模型的正确性,利用训练好的神经网络模型,采用调整后的权值和阈值,将2010年的数据作为独立样本进行预测检验。结果表明:基于BP神经网络的水质指标预测模型收敛速度快,对训练样本具有很好的拟合能力,且对检验样本的预测精度较高。

By taking the Huajiangyingzi section weekly quality index data from 2004 to 2009 as the training samples,prediction models are trained and established respectively based on BP neural network of pH,dissolved oxygen,ammonia nitrogen and permanganate index.In order to verify the correctness,the neural network models are trained with adjusted weights and thresholds,and the data in 2010 are used as an independent test sample.The results show that the BP prediction models of water quality indexes have fast convergence speed,good fitting capability to the training samples,and high forecast accuracy to the test sample.