摘要
为更好地反映测站水位流量关系所固有的规律,提高水位流量关系模拟的稳定性和精度,分析了影响单一线法定线与推流精度的几种因素,在采用门限回归建模、BP网络建模的基础上,分析探讨了将门限回归与BP网络结合起来的方法:先用门限回归模型将实测样本分段,再用BP网络对各分段样本进行非线性拟合。实例研究表明:该方法克服了单独使用门限回归方法因分段数不足而无法较好地模拟总体的非线性规律,以及当实测样本存在突变和分段规律时BP网络方法本身无法对其甄别的缺陷,较其他两种方法精度更高、效果更好,丰富了目前现有的稳定水位流量关系分析的单一线法,可在工程实践中尝试应用。
In order to reflex preferably the inherent laws of stage discharge relation at a station and improve the simulation stability and precision of the stage discharge relation, this paper analyzed several factors that affected the single curve method's determination and flow derivation. On the basis of using threshold regression and BP net for modeling, a new method was discussed based on threshold regression and BP net: using threshold regression model to divide the measured sample into several segments first, then using BP network to execute nonlinear fitting the segments. The case study shows that the new method overcomes the disadvantages using separately threshold regression to simulate nonlinear rules of population when segments are insufficient, and the defects that BP network itself doesn't identify when the measured samples exist mutation and the section rule. The new method has higher accuracy and better stability than the other two methods, which enriches the existing single curve method on stable stage discharge relation analysis, and it can be applied in engineering practices.
作者
程银才
王军
范世香
CHENG Yincai WANG Jun FAN Shixiang(Water Conservancy and Civil Engineering College, Shandong Agricultural University, Taian 271018, China Hydrology Bureau of Taian, Taian 271000, China)
出处
《人民黄河》
CAS
北大核心
2016年第6期55-57,62,共4页
Yellow River
关键词
门限回归
BP网络
水位流量关系
定线
推流
threshold regression
BP net
stage discharge relationship
determination of relation curve
derivation of flow