摘要
为验证BP神经网络模型在预测直角折线堰过流能力方面的适用性,利用量化共轭梯度算法,选取114组水工试验数据为训练样本、21组水工试验数据为预测样本,将直角折线堰前堰长度(a)、侧堰长度(b)及堰顶水头(H)作为输入层神经元,流量Q作为输出层神经元,建立3-10-1的拓扑结构直角折线堰过流能力预测模型。结果表明,基于BP神经网络的模型预测值与试验值的最大误差仅为1.76%,相关系数达0.9997。基于验证后的预测模型,给出了直角折线堰的流量系数表,可为工程设计提供参考。
To verify the applicability of BP neural network in prediction of the right angle polyline weir's overflow capacity,using the quantized conjugate gradient algorithm,114sets of data in the experimental research results were selected as training samples and 21groups of data were used as prediction samples.The front weir length(a),lateral weir length(b)and water head of weir top(H)were chosen as input layer neurons,and flow rate Q was taken as output layer neurons.A 3-10-1topological structure right-angle polyline weir overflow capability prediction model was established.The results show that the maximum error between the model prediction value and the experimental value by the BP neural network is 1.76%,and the correlation coefficient is 0.9997.Based on the verified prediction model,the discharge coefficient table of the right angle polyline weir was given,which can provide a reference for the design of engineering.
作者
邱勇
杨泽文
周鑫宇
吴锦钢
谢红英
QIU Yong;YANG Ze-wen;ZHOU Xin-yu;WU Jin-gang;XIE Hong-ying(College of Water Resources and Hydraulic Engineering,Yunnan Agricultural University,Kunming 650201,China)
出处
《水电能源科学》
北大核心
2021年第3期74-77,共4页
Water Resources and Power
基金
国家级大学生创新创业训练计划项目(201810676004)。
关键词
流量系数
过流能力
预测模型
BP神经网络
直角折线堰
discharge coefficient
overflow capability
prediction model
BP neural network
right-angle polyline weir
