基于BP神经网络的氢气传感器数据拟合与研究  

Research on hydrogen sensor data fitting based on BP neural network

作　　者：李颖 牛萍娟[1] 刘宝丹[2] LI Ying;NIU Pingjuan;LIU Baodan(School of Electrical Engineering and Automation,Tianjin Polytechnic University,Tianjin 300387,China;Institute of Metal Research,Chinese Academy of Sciences,Shenyang 110016,China)

机构地区：[1]天津工业大学电气工程与自动化学院,天津300387 [2]中国科学院金属研究所,辽宁沈阳110016

出　　处：《现代电子技术》2021年第24期97-101,共5页Modern Electronics Technique

基　　金：天津市科技计划项目(18ZXZNGX00130)。

摘　　要：随着氢能源的推广,对高精度的氢气传感器需求日趋强烈。为建立氢气传感器测量模型,简化传统传感器测量电路,文中从传感器的基本检测电路入手,推导传感器等效电阻、功耗和灵敏度系数的表达式。首先拟合得到氢气浓度与灵敏度系数的回归方程,但测试样本的拟合优度欠佳;为获得更为精确的模型,引入BP神经网络训练算法,建立传感器输出响应与气体浓度的网络模型。仿真中预测样本的平均误差降至9.74%,决定系数R2高达0.99962。相较于拟合样本,测试样本的相对误差减小了19.54%,R2提升了2.14%,表明BP神经网络对氢气浓度具有更好的预测效果。同时,文中的研究为氢气传感器的数据分析提供了一种更优的方法。With the rapid promotion of hydrogen energy,the demand for high⁃precision hydrogen sensors is becoming stronger.In order to establish a hydrogen sensor measurement model and simplify the measurement circuit of the traditional sensor,the expressions of the sensor′s equivalent resistance,power consumption and sensitivity coefficient are derived in starting from analysis of the basic detection circuit of hydrogen sensor.The regression equation of the hydrogen concentration and the sensitivity coefficient is obtained by fitting,but the goodness of fitting is not good.To obtain a more accurate model,BP training algorithm is introduced,and a network model of sensor output response and gas concentration is established.The average error of the prediction samples in the simulation is reduced to 9.74%,and the coefficient of determination is as high as 0.99962.In comparison with the fitting samples,the relative error of BP′s prediction value of hydrogen concentration is decreased by 19.54%,and R2 is increased by 2.14%,which indicate that BP neural network has a better prediction effect on gas concentration.The research in this paper provides a better method for the data analysis of the hydrogen sensor.

关 键 词：数据拟合 氢气传感器 BP神经网络 测量模型 氢能源检测 相对误差 

分 类 号：TN245-34[电子电信—物理电子学] TP212.9[自动化与计算机技术—检测技术与自动化装置]