IQR-MAD-Based Anomaly Detection of Voltage Data in the Distribution Network
Bo Fu, Yang Liu, Ying-Dan Wang, Hao Peng, Yuanli Chen, Yu Hua Cao · 2023
Due to the influence of the natural environment, measuring instruments and other factors, the anomalous measurement data will lead the dispatchers to make wrong decisions, threatening the safe and stable operation of the power system. To improve the speed and accuracy of identifying anomalous voltage data of distribution network voltage monitoring points, and avoid the influence of anomalous values on statistical parameters (mean value and standard deviation), a two-layer progressive anomaly detection method for distribution network voltage data based on interquartile range (IQR) and mean absolute deviation (MAD) is proposed. In the first layer, the voltage data collected by the distribution network voltage monitoring system is firstly filtered with zero value and null value, and then the significant outliers are detected according to the IQR criterion to obtain the data set to be further detected. The second layer proposes the MAD-based Hampel algorithm as an outlier evaluation criterion, and the anomaly judgment coefficient can be adjusted according to the needs of power grid operators. Finally, in order to evaluate the proposed method, two real distribution network voltage data are taken as examples to compare traditional statistics-based anomaly detection algorithms, such as the Three Sigma Rule and IQR, through simulation. The results show that the proposed algorithm has a higher accuracy rate than the comparison algorithm.