摘要
为掌握钱塘江水质未来的变化情况以及预防污染事件的发生,建立了一个水质指标预测模型.利用钱塘江某行政交界断面的水质指标实测数据作为学习样本,选取了总磷、总氮、化学需氧量等9项指标作为预测参数,运用Levenberg-Marguardt优化算法对学习样本进行优化,建立了反向传播(BP)神经网络模型,并运用该模型对钱塘江水质指标进行了预测.结果表明,BP神经网络模型的预测精度较高,预测速度快,对大部分水质指标能够得到较好的预测值,相对误差的绝对值小于6%.此BP神经网络能够有效地应用于水质指标的预测和水质趋势的预警预报系统中.
A predictive model was set up to obtain clear knowledge of the prospective changing conditions of water quality in Qiantang River and prevent further pollution. The historical time series of water quality indexes in district border reaches of Qiantang River were taken as instructive samples, and nine indexes were taken as predicted indexes, such as total phosphor, total nitrogen, chemical oxygen demand. The samples were modeled and optimized with Levenberg-Marguardt algorithm of back propagation (BP) neural network. The predicted results indicated that the prediction of water quality was precise and fast, and that the relative errors of the predicated indexes, with a few exceptions, were lower than 6 %. The BP neural network can be satisfactorily applied to predict water quality index and is suitable for pre-alarm of water quality trend.
出处
《浙江大学学报(工学版)》
EI
CAS
CSCD
北大核心
2007年第2期361-364,共4页
Journal of Zhejiang University:Engineering Science