Anomaly Detection in Power Grid IoT System based on Isolated Forest

Abstract

Along with the development of IoT (Internet of Things) infrastructure in power grid, data quality is becoming essential for the management and safety of electric power grid system. In this paper, a data anomaly detection framework is proposed based on isolated forest. This algorithm first improves the feature selection by calculating the Kurt factor. Then, Local sensitive hash for spatial mapping is applied to create the sub-feature set for the isolated forest. The proposed method is tested on the IoT datasets (temperature, humidity, CO2 and etc.) from the power distribution stations. The experimental results indicate that the proposed algorithm achieves better performance in terms of accuracy and recall than existing methods such as Kmeans, SVM or standard isolated forest.