Qiaoyun Liu, Huifeng Xue, Shaopeng Zhou, and Han Wang. Abnormal large-valued gradient-frequency compound rejection methodfor water resources monitoring data. Dynamic Systems and Applications 29 (2020) No. 4, 1622 – 1636

https://doi.org/10.46719/dsa202029445

ABSTRACT.
Under the background of the construction of Digital China, the monitoring data of water resources is expanding massively, but there are various factors in the actual monitoring data that lead to abnormally large values, which seriously affect the correctness and credibility of the statistical analysis of water resources. To solve this problem, based on the analysis of the large value of the anomaly and its distribution characteristics, the gradient frequency composite elimination method (GFCR) is proposed. According to the left and right derivatives and the second derivatives of discrete data series, the criterion of outliers is designed, the threshold is determined by the frequency distribution of data set, the criterion of outliers rejection is given, and the algorithm of outliers gradient frequency rejection is designed. GFCR method is used to process the daily water intake time series data of four monitoring points in a given region. According to the experimental results, 64 data were removed from 4 nodes, and the total of the removed abnormal values accounted for 24.40% of the total data, which ensured the correctness of the monitoring data statistics and laid a foundation for the correct analysis of water resources.

Keywords: Water resource; Monitoring data; Abnormal large values; Data rejection; Gradient-frequency composition; Statistical analysis