【作者】 刘载文； 崔莉凤； 王小艺； 吕思颖；

【Author】 Liu Zaiwen,Cui Lifeng,Wang Xiaoyi,Lu Siying Beijing Technology and Business University,Beijing 100037,P.R.China

【机构】 北京工商大学信息工程学院；

【摘要】 分析水华发生的主要因素,研究水华的软测量主导变量确定方法与短期预测模型,RBF神经网络基函数中心、宽度和权值的监督学习方法,基于梯度下降的误差纠正算法,提出基于RBF网络的水华软测量方法。对RBF网络隐层节点数对网络性能的作用和影响进行了分析比较,对径向基函数宽度与网络拟合能力,以及网络泛化性能进行讨论。从网络的训练和水华预测结果来看,采用RBF神经网络可以预测叶绿素的短期变化规律,基于RBF网络的水华软测量模型泛化能力强,网络预测精度高,拟合性能好,为河湖水华短期预测方法的深入研究奠定了基础。更多还原

【Abstract】 After the major elements of water bloom were analyzed,a method to confirm dominant variables of soft sensing and shortdated forecast model for water bloom,and a supervise learning method to determine the clustering center,extent and weighting function value of radial base function were studied,then the soft sensing method based on RBF neural network for water bloom was proposed,The fitting ability and extensive capability of neural network are discussed according to different nodes of hidden layer and the extent of RBF neural network.The results of neural network training and water bloom forecast show that the soft sensing model based on RBF neural network possess strong extensive capability and forecast precision,and it can provide a new method and research groundwork for the forecast of water bloom in river and lakes.更多还原