基于EMD-ABOD的大坝异常监测数据识别方法研究

Research on Identification Method of Anomalous Monitoring Data of Dams Based on EMD-ABOD

大坝监测数据普遍存在异常值,对异常数据进行识别和剔除,可保持模型的稳定性和可靠性,并提高模型的预测或分类性能;同时,可及时发现异常情况,以保证系统的安全运行。因此,将基于角度的异常值检测算法(ABOD)引入大坝监测异常数据识别,首先通过经验模态分解(EMD)提取监测数据的高频本征函数,然后对由高频本征函数构成的新数据进行异常数据识别。对长河坝沉降监测数据的验证结果表明,与其他方法相比,EMD-ABOD可有效提升异常数据识别的准确性。

There widely exists anomalous data in dam monitoring.Identifying and removing anomalous data contributes to maintaining the stability and reliability of models,thereby enhancing their predictive or classification performance.At the same time,real-time monitoring and anomaly detection can ensure the safe operation of systems.This paper introduces the Angle-Based Outlier Detection(ABOD)algorithm for identifying anomalous data in dam monitoring.Firstly,Empirical Mode Decomposition(EMD)was used to extract the high-frequency intrinsic mode functions of monitoring data.Subsequently,anomaly detection is performed on the new dataset composed of these high-frequency intrinsic mode functions.Applied to data from Changheba,through comparative analysis with other methods,EMD-ABOD demonstrates an effective enhancement in the accuracy of anomalous data identification.