基于BP神经网络的人工加糙明渠糙率预测模型  

Prediction Model of Roughness Coefficient of Artificially Roughened Open Channel Based on BP Neural Network

作　　者：王麟 赵涛[2] WANG Lin;ZHAO Tao(College of Hydraulic and Civil Engincering,Xinjiang Agricultural University,Urumqi 830052,China;Xinjiang Key Laboratory of Hydraulic Engineering Security and Water Disasters Prevention,Xinjiang Agricultural University,Urumqi 830052,China)

机构地区：[1]新疆农业大学水利与土木工程学院,新疆乌鲁木齐830052 [2]新疆农业大学新疆水利工程安全与水灾害防治重点实验室,新疆乌鲁木齐830052

出　　处：《水电能源科学》2021年第9期125-127,124,共4页Water Resources and Power

基　　金：新疆维吾尔自治区自然科学基金项目(2015211A025)。

摘　　要：糙率是明渠水力计算的各项影响因素中最重要的参数。运用BP神经网络的方法,结合其在结构上的稳定性及在处理非线性数据上的优势,以矩形人工加糙明渠为研究对象,建立预测模型研究其各水力要素与糙率间的复杂非线性关系。根据前期的试验研究成果,选择绝对粗糙度Δ、底坡i、弗劳德数Fr、流量Q作为主要影响因素对糙率进行神经网络建模及预测,并将预测结果与径向基(RBF)神经网络及偏最小二乘及最小二乘支持向量机(PLS-LSSVM)进行对比。研究结果表明,基于L-M算法的BP神经网络糙率预测模型的平均绝对百分比误差为0.51%,均方根误差为8.15×10-5,精度优于其他预测模型,说明BP神经网络可有效预测矩形人工加糙明渠的糙率。The roughness coefficient is the most important parameter for influencing hydraulic computation of open channel.Back-propagation(BP)neural network method has advantages of structural stability and processing nonlinear data.Taking the rectangular artificially roughened open channel as the research object,aprediction model was established to study the complex nonlinear relationship between hydraulic factors and roughness coefficient.According to previous test research results,absolute roughnessΔ,bottom slope i,Froude number Fr and inflow discharge Q were selected as the main influencing factors to simulate and forecast roughness by using neural network.Radial basis function(RBF)neural network and partial least square-least square support vector machine(PLS-LSSVM)were compared with the predicted results of the neural network simulation.The results show that the roughness coefficient prediction simulation by the BP neural network based L-M algorithm is more accurate than RBF neural network and PLS-LSSVM.The average absolute percentage error(MAPE)and root mean square error(RMSE)of BP neural network are 0.51%and 8.15×10-5 respectively.It is indicated that BP neural network can effectively predict the roughness coefficient of rectangular artificially roughed open channels.

关 键 词：BP神经网络 人工加糙明渠 预测 糙率系数 

分 类 号：TV133[水利工程—水力学及河流动力学]