摘要
引黄灌渠斗口水流量通常依据闸门开度、闸前和闸后水位等可观测信息估算。自动测量装置具有“软仪表”的典型特征,建立精确适用的软测量模型十分关键。鉴于训练后的人工神经网络可以精确逼近任意非线性函数,建立了基于BP网络和RBF网络的引黄灌渠斗口流量软测量模型,并精选水工试验数据构成训练样本集进行仿真训练。检验表明,基于人工神经网络的软测量模型输出值与期望值(标准三角量水堰的测量结果)吻合良好,斗口水流量软测量精度有显著改善。
Water discharge at the outlet of irrigation channel in the Yellow River is estimated through measurable information such as open-level of floodgate, water levels at front and back of the floodgate and so on. The auto-measuring device has typical characteristics of soft-meter, therefore, it is most important to establish an exact model. Because trained ANN can exactly approach to any nonlinear function, ANN-based soft-sensing models (BP and RBF) are established and trained by sample-set selected from experiment-data. The test shows that outputs of the ANN-based models approach to expected value (result from standard triangular weir) well. The precision of soft-sensing for discharge at the outlet of irrigation channel is improved remarkably.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2005年第10期994-997,共4页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(60165001)资助项目
关键词
引黄灌渠
流量测量
软测量
ANN
Irrigation channel in the Yellow River
Discharge measurement
Soft-sensing
ANN