朱 詠1，陳學林2

(1.甘肅省張掖水文水資源勘測局，甘肅 張掖 734000;2.甘肅省水文水資源局，甘肅 蘭州 730000)

　　摘 要：分析天然河道糙率值的主要影響因素，基于LM-BP神經網絡預測模型，提出黑河鶯落峽水文站測驗河段糙率值的推求方法。對網絡模型的設計、訓練和測試進行研究和分析，預測模型輸入量為水位Z、水力半徑R及水面比降S，輸出量為糙率n，樣本集為黑河鶯落峽水文站測驗河段的歷史實測糙率值。根據測試結果可以得出，在誤差允許范圍內，采用LM-BP神經網絡預測模型可較準確地推求不同水位、比降對應的n值,該糙率值可應用到河道高洪水期比降面積測流中。

　　關鍵詞：糙率;LM-BP神經網絡;天然河道;推測

　　The speculating research of friction in the Yingluoxia channel based on LM-BP neural network

　　ZHU Yong1, CHEN Xuelin2

　　(1.Zhangye Hydrology and Water Resources Bureau of Gansu Province, Zhangye 734000, China;

　　2. Hydrology and Water Resources Bureau of Gansu Province, Lanzhou 73000,China)

　　Abstract: The article analyzes the main influencing factors in the process of natural channel roughness value. Based on the LM-BP neural network prediction model, it proposes a method to calculate the roughness value of the tested reach of the Yingluoxia hydrological station in Heihe River, researches and analyzes the design, training and testing of network model. The input of the prediction model is water level z, hydraulic radius R and water surface slope S; the output is roughness n. The sample set is the historical measured roughness value of the tested reach of Yingluoxia Hydrometric Station in Heihe River. According to the test results, within the allowable error range, the n value corresponding to different water level and specific water surface slope can be accurately calculated by LM-BP neural network prediction model. The n value can be applied to the flow measurement of slope-area in high flood period.

　　Key words: roughness; LM-BP neural network; natural channel; speculation