基于BP神经网络的海河干流叶绿素浓度短时预测研究

发布时间：2018-03-17 16:32

本文选题：海河干流　切入点：叶绿素浓度　出处：《水利水电技术》2017年11期 　论文类型：期刊论文

【摘要】：为实时预测海河干流水体藻华的暴发时段及影响程度,提高环境管理部门决策能力,以海河干流段典型断面的水质在线监测及气象站高频、实时数据为基础,基于BP神经网络,以实时叶绿素浓度、气温、光照强度和气压四项指标为输入变量,建立了叶绿素浓度日变化量的预测模型,对海河干流大光明桥处水域叶绿素浓度随时间的变化进行预测。结果表明:对海河干流叶绿素浓度短时预测影响较大的因素依次为溶解氧(叶绿素)、气温、光照强度、气压、降雨、电导率、相对湿度;预测时长越短,预测精度越高。当预测时长分别为24 h、12 h、6 h时,Nash效率系数分别为0.77、0.85、0.93,预报误差的标准误差分别为5.7μg/L、4.6μg/L、3.1μg/L;12 h内的预测精度可满足海河河道藻华预警的实际需求,为其短期预警提供了数据支撑。
[Abstract]:In order to predict the outbreak period and influence degree of algal blooms in Haihe River main stream in real time, and improve the decision-making ability of environmental management department, based on the on-line monitoring of water quality of typical sections of Haihe River and the high frequency and real time data of meteorological stations, BP neural network was used. A prediction model of diurnal variation of chlorophyll concentration was established with real-time chlorophyll concentration, temperature, light intensity and pressure as input variables. The variation of chlorophyll concentration with time at Daguang Bridge of main stream of Haihe River is predicted. The results show that the factors that influence the short-term prediction of chlorophyll concentration in the main stream of Haihe River are dissolved oxygen (chlorophyll), temperature, light intensity, pressure, etc. Rainfall, conductivity, relative humidity; shorter prediction time, The higher the prediction accuracy is, the higher the prediction accuracy is, when the prediction time is 24 h / 12 h / 6 h respectively, the Nash efficiency coefficient is 0.77 / 0. 85 / 0. 93 and the standard error of prediction error is 5. 7 渭 g / L / L ~ (4.6) 渭 g / L ~ (-1) / L ~ (-1) 3 渭 g / L ~ (-1) / L ~ (-1) / L ~ (12) h respectively, which can meet the actual demand for early warning of algal blooms in Haihe River. For its short-term warning to provide data support.
【作者单位】：天津大学;天津市环境监测中心;天津市环境保护科学研究院;三峡大学;
【基金】：国家“十二五”水体污染控制与治理科技重大专项“海河干流水环境质量改善关键技术与综合示范”(2014ZX07203-009)
【分类号】：TP183;X522

【相似文献】