基于鲁棒估计的动态传感数据校正方法  

Nonlinear Dynamic Data Reconciliation Method Based on Robust Estimation

作　　者：黄国健[1] 刘桂雄[1] 洪晓斌[1] 高屹[1] 

机构地区：[1]华南理工大学机械与汽车工程学院

出　　处：《自动化与信息工程》2009年第4期6-9,共4页Automation & Information Engineering

基　　金：广东省教育部产学研结合项目(编号2007A090302039);教育部新世纪优秀人才支持计划项目(NCET-08-0211)资助

摘　　要：为实现IEEE1451网络化智能传感器输出数据与输入信号随时间的变化曲线保持一致,解决智能传感中的非线性问题,根据数据校正与数据回归的相似性,引入并改进在数据回归中应用较好鲁棒估计方法,结合非线性动态数据校正的DAEs模型,为传感器的校正引擎构建了相应的鲁棒估计目标函数。应用实例可得,校正后的结果与模型值的均方差MSE<0.003,表明在数据校正方法中引入鲁棒估计方法,可以使目标函数对偏离理想条件的数据不敏感,能显著减小异常数据点对传感数据校正的不利影响,实现高效的数据自校正。To correct the IEEE 1451 networked intelligent sensor＇s output data and input signal curve, solve the nonlinear dynamic data reconciliation （NDDR） problem, this paper introduces robust estimation method to NDDR because of the resemblance between data correction and data regression. Combining with the differential algebraic equations model of NDDR, an objective function based on robust estimation was constructed for the correction engine of an intelligent sensor. Research results show that the MSE between the corrected data and model data is less than 0.003, and the objective funetion of robust estimation is not sensitive to the deviation data from the ideal conditions, which can significantly reduce the abnormal data＇s adverse effects in sensor data correction and achieve efficient data self-correction.

关 键 词：鲁棒估计 非线性 数据校正 智能传感器 

分 类 号：TP212.6[自动化与计算机技术—检测技术与自动化装置] TP391.41[自动化与计算机技术—控制科学与工程]