摘要
河川径流等水文时间序列属于复杂的非线性系统, 使用回归分析等传统的分析方法, 难以获取和描述其内在关联和变化规律。利用现有的相关站点的径流量历史数据和输沙量、降水量数据, 在进行规格化处理和主成分分析的基础上, 利用三层BP人工神经网络模型, 对澜沧江流域上游昌都站径流量与各关联因子之间复杂的非线性映射关系进行模拟, 采用拟牛顿算法对模型进行训练, 模拟结果达到期望精度要求, 并利用1982年~1985年实测数据进行模型验证。结果证明利用BP神经网络模型对澜沧江流域站点的月径流量序列进行模拟、预测和数据补缺处理具有可行性。
The time series of hydrology were complicated non-linear system, such as the runoff of a river. It was difficult to obtain or discribe the internal relationship and the law of the variation. A BP( back propagetion) aritificial neural network model was used to express the complicated non-linear relationship between the runoff of Changdu and other interrelated factors after normalizing the inputs and targets and analysis the principal component based on the runoff, precipitation, and data we had. An imitating Newtonian algorithm was used in the train of the model. The precision of simulating result of the model reached the expectation. The model was validated with the real data of year 1982 - 1985. It was proved to be feasible that the BP artificial neural network could be used in the simulating, forecasting and vacancy filling of the series of the runoff of the Lancang River Basin.
出处
《测绘科学技术学报》
北大核心
2008年第4期271-274,共4页
Journal of Geomatics Science and Technology
基金
科技部社会公益研究专项--<澜沧江国际河流水资源环境空间分析平台>(2005DIB3J160)项目资助
关键词
人工神经网络模型
规格化
主成分分析
径流量序列
模拟补缺
artificial neural network
normalizing
the Lancang river basin
runoff series
vacancy filling