雨量检定仪基于迁移学习的自标定算法  

Self calibration algorithm for rainfall gauge based on transfer learning

作　　者：孔令旺[1] 党选发[1] 第五朋朋 韩玉婷[1] 陈小贻 KONG Lingwang;DANG Xuanfa;DIWU Pengpeng;HAN Yuting;CHEN Xiaoyi(Meteorological Information and Technology&Equip Support Center of Gansu,Lanzhou 730020,China)

机构地区：[1]甘肃省气象信息与技术装备保障中心,兰州730020

出　　处：《国外电子测量技术》2025年第1期134-139,共6页Foreign Electronic Measurement Technology

基　　金：甘肃省科技计划(24JRRA1185)。

摘　　要：针对雨量检定仪的出水流速与控制脉冲频率呈非线性关系,且不同雨量检定仪在同一控制脉冲频率下流速存在差异的问题,设计了雨量检定仪基于迁移学习的自标定流程和算法。先利用检定仪流速和驱动频率的物理关系式,对两者关系进行神经网络预训练,得到预训练结果。再利用雨量检定仪自带的电子元器件,通过不同脉冲频率控制雨量检定仪多次进水和出水,采样检定仪控制频率和出水速率的实际对应关系,最后使用预训练结果对采样点数据进行迁移学习,获得检定仪控制频率和出水速率之间的实际关系,实现了雨量检定仪的自标定。实验表明,该流程和算法在无需其他辅助标定设备的条件下,对检定仪进行了自标定,提高了雨量自动检定仪的稳定性和精度。In view of the problem that the water outflow velocity of the rain gauge calibrator has a non-linear relationship with the control pulse frequency,and there are differences in the flow velocity of different rain gauge calibrators at the same control pulse frequency,a self-calibration process and algorithm for the rain gauge calibrator based on transfer learning was designed.First,the physical relationship between the flow velocity of the calibrator and the driving frequency was used to pre-train the relationship between the two through a neural network,obtaining the pre-training results.Then,the electronic components built into the rain gauge calibrator were utilized to control the calibrator to intake and outflow water multiple times at different pulse frequencies,and the actual corresponding relationship between the control frequency of the calibrator and the water outflow rate was sampled.Finally,the pre-training results were used to perform transfer learning on the sampled data points,obtaining the actual relationship between the control frequency and the water outflow rate of the calibrator,thus achieving the self-calibration of the rain gauge calibrator.Experiments show that this process and algorithm can self-calibrate the calibrator without the need for other auxiliary calibration equipment,improving the stability and accuracy of the automatic rain gauge calibrator.

关 键 词：雨量 检定仪 蠕动泵 迁移学习 自标定算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] P414[自动化与计算机技术—计算机科学与技术]