考虑时间序列关联的大坝监测异常数据清洗

针对传统大坝安全监测中异常数据清洗方法主要分析单一监测效应量,较难区分粗差和环境突变引起的异常值,提出一种利用关联规则约束和引导大坝安全监测异常数据清洗方法。首先,通过关联规则辨识强关联性序列,结合基于密度的聚类算法识别序列中异常数据;然后,根据关联序列异常数据清洗规则,辨识大坝安全监测中粗差数据,并利用基于粒子群优化最小二乘支持向量机模型重构异常数据。最后,对大坝典型位移量数据进行异常数据清洗。结果表明:该方法能够甄别监测效应量中环境突变引起的异常值,提高了大坝安全监测数据中数据清洗的准确性。

Cleaning of Abnormal Dam Monitoring Data Considering Time Series Association

Because the traditional abnormal data cleaning method in dam safety monitoring mainly analyses the single monitoring effect,it is difficult to distinguish the abnormal values caused by gross errors and environmental sudden changes.Aiming at this problem,a new method of abnormal data cleaning for dam safety monitoring is proposed herein,which uses association rules to restrict and guide the abnormal data cleaning.Firstly,the strong correlation sequence is identified by association rules,and the abnormal data in sequence is identified by density-based clustering algorithm.Secondly,the gross error data in dam safety monitoring is identified by association sequence abnormal data cleaning rules,and the abnormal data is reconstructed by particle swarm optimization least squares support vector machine model.Finally,the abnormal data cleaning of dam typical displacement data is conducted.The result shows that the proposed method can discriminate the abnormal values caused by environmental sudden change in monitoring effect,and improve the accuracy of data cleaning for dam safety monitoring data.