大坝位移监测中的粗差剔除方法  

A Gross Error Elimination Method for Dam Displacement Monitoring

作　　者：花胜强[1] 胡少英[1] 罗孝兵[1] 高磊[1] 郑健兵[1] 

机构地区：[1]南瑞集团公司(国网电力科学研究院),江苏南京211106

出　　处：《水力发电》2017年第9期110-112,共3页Water Power

摘　　要：针对大坝位移监测中常用粗差剔除法存在的问题提出了一种新型的粗差剔除方法。首先,根据层次分析法得到Dixon准则、Grubbs准则、Chauvenet准则、四分点准则的权重系数,并设定好粗差判定阈值;其次,根据原始观测资料建立大坝位移的偏最小二乘回归模型,得出样本对应的残差序列;然后根据四个准则依次判定出所有疑似粗差的测值;最后,按照此4种准则的权重系数,加权综合得到每个测值的粗差概率值,并与粗差判定阈值对比,得到最后的粗差集并剔除。该方法综合考虑了多种粗差数据识别方法,权衡了误判和漏判的可能性,为剔除大坝位移数据中的粗差提供了一种更为科学合理的方法。A new method for gross error elimination of dam displacement monitoring data is proposed. Firstly, the weight coefficients of Dixon criterion, Grubbs criterion, Chauvenet criterion and four-point criterion are obtained according to analytic hierarchy process, and the threshold of gross error judgment is set. Secondly, the partial least squares regression model of dam displacement data is established based on original observation data and corresponding residual sequence is obtained. Thirdly, all the suspected gross errors are determined by the four criterions. Finally, the gross error set is gotten and eliminated by the weighted sum of gross error probability of each criterion result compared to the threshold. This method comprehensively considers a variety of gross error data recognition methods, and weighs the possibility of miscarriage of justice and missed judgment, that provides a more scientific and reasonable method to eliminate the gross error of dam displacement monitoring data.

关 键 词：大坝位移监测 粗差剔除 Dixon准则 Grubbs准则 Chauvenet准则 四分点准则 

分 类 号：TP698.1[自动化与计算机技术—控制理论与控制工程]