摘要
针对实时在线递推参数对含粗差的观测量相当敏感,个别粗差就会对参数的估值产生较大的影响的这种情况,本文通过选择合适的权函数,采用抗差最小二乘来代替传统最小二乘法来在线估计AR模型参数。对鄂北石油钻井数据的大钩负荷测量数据进行仿真,对抗差最小二乘和传统最小二乘法结果进行比较,结果表明,抗差递推最小二乘法能得到满意的参数估计,同时能满足该系统的实时性,具有更强的容差能力。
The real-time AR model parameters are sensitive to gross error and easily to be influenced. The paper uses the robustified recursive least square method instead of the standard reeursive least square method to estimate the parameters on line. We simulate the hookweight observations of the oil well in E-bei by using the two different methods mentioned above. The comparison of two sim- ulation results proves that the robustified reeursive least square method can get satisfied parameter estimation and meet the system＇ s dynamic needs. So we can say that it has much better performance in dealing with outliers.
出处
《微计算机信息》
北大核心
2008年第30期193-195,共3页
Control & Automation
基金
河南省杰出人才创新基金项目:智能化网络入侵防御系统关键技术研究(074200510013)
河南省教育厅自然科学基金项目:网络协议兼容和规则完备的智能入侵防御系统研究(2007520048)
关键词
抗差最小二乘法
权函数
AR模型
粗差
实时在线递推
robustifled recursive least square method
weight function
AR model
gross error
real-time recursion
