基于LOF与CEEMD的城镇取用水监测数据异常值识别

为有效识别城镇取用水监测数据异常值,提高数据的可靠性与真实性,结合局部异常因子(LOF)算法与互补集成经验模态分解(CEEMD)法,开发城镇取用水监测数据异常值自动识别的方法.先应用LOF进行可直观异常值识别,再应用CEEMD对修正后的数据序列进行频谱分解,通过低频叠加分量拟合序列并设定相对误差阈值用以识别不可直观异常...

Outlier Identification of Urban Water Intake Monitoring Data Based on LOF and CEEMD

In order to identify the outliers of urban water intake monitoring data effectively and improve the reliability of the data, the automatic outlier identification method is developed by combining the Local Outlier Factor (LOF) method with the Complementary Ensemble Empirical Mode Decomposition (CEEMD) method. LOF is used to identify observable outliers firstly and then CEEMD is applied for spectral decomposition of the revised data series. Sequences are fitted by low-frequency superposition components, and the relative error threshold is set to identify non-observable outliers. Taking the monitoring data of daily water intake of a waterworks in Hebei Province for experimental analysis, and the results show that the revised annual water intake data reduces from 512 700 m³ to 411 400 m³ . The revised data is much closer to the manually approved annual data. And therefore there is a large error if the monitoring data was used directly to calculate the total annual water intake and consumption. The proposed method can effectively identify and correct the outliers in the urban water intake monitoring data, and provide technical support for the follow-up strong supervision of water resources.