基于神经网络的数据融合在废气测量中的应用  

Application of Data Fusion to Measurement of Exhaust Gases Based on Neural Network

作　　者：朱伟兴[1] 李丽[1] 庞敏[1] 

机构地区：[1]江苏大学电气信息工程学院

出　　处：《中国安全科学学报》2007年第6期162-165,共4页China Safety Science Journal

基　　金：江苏省国际合作项目(BZ2005035);江苏省镇江市农业科技项目(NY2006047)

摘　　要：数据融合方法通过提取各个影响因素之间的特征关系,进行数据之间的融合。针对因传感器故障而失真的数据,综合考虑对畜禽场排放的某一废气测量值的时间、空间和环境等多种影响因素,使用基于神经网络的数据融合方法来估算该废气的浓度,实现失真数据的恢复,从而精确地测量出养殖场连续排放的有害气体的总量,对超标排放进行监控。以氨气(NH3)浓度数据的处理为例,应用MATLAB软件,其仿真结果表明:估算最大相对误差为7.83%,证明基于神经网络的数据融合方法的有效性。The method of data fusion fulfills the fusion among data by extracting the characteristic relation between influencing factors. Directing at the data distortion caused by the fault of sensors, and considering the factors influencing the measurement of certain exhaust gas in an animal barn such as temporal, spatial and environmental parameters, neural network is introduced to data fusion so as to figure out the missed data due to the faults of this sensor measuring the value of this exhaust gas. By this method, total emission of nocuous gases from the barn is measured accurately, and the emission of exhaust gases beyond normal standard was controlled. Taking the concentration data of NH3 for example, MATLAB simulation results show that the maximal relative error of estimation is 7.83%. Thus, it is verified that this method of data fusion based on GRNN （Generalized Regression Neural Network） is feasible and valid.

关 键 词：广义回归神经网络(GRNN) 数据融合 禽畜养殖场 废气 缺失数据 

分 类 号：X701[环境科学与工程—环境工程] TP183[自动化与计算机技术—控制理论与控制工程]