Open AccessMethod for Detecting Anomaly Data of WAMS System Based on GA-iForest
Author(s) -
Hongwen Yan,
Jiawei Li,
Jian Hua Zuo,
Jihong Tang
Publication year - 2020
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/512/1/012161
Subject(s) - anomaly detection , stability (learning theory) , data mining , computer science , anomaly (physics) , power grid , electric power system , genetic algorithm , grid , power (physics) , mathematics , machine learning , physics , geometry , quantum mechanics , condensed matter physics
Wide area measurement system problems such as anomalies, data missing, and real-time response deeply affect the operation and maintenance of the power grid. Considering the accuracy requirements of the system for anomaly detection, the detection accuracy and stability of traditional isolation forest are poor, so this paper proposes a new data anomaly detection method named GA-iForest. This method uses genetic algorithms to select isolated trees with high accuracy and obvious differences to optimize the structure of isolated forests. The new data anomaly detection method GA-iForest improves the accuracy and stability of detection, and mean while maintaining the high efficiency. The experiment uses standard simulation data sets and data from one province’s wide-area measurement system for experimental simulation. The result shows that the GA-iForest method has significantly improved accuracy and stability compared with the traditional isolation forest, LOF, and K-means methods.