Anomaly detection in smart grid using optimized extreme gradient boosting with SCADA system

The significance of anomaly detection is paramount for ensuring the security and better cost-efficiency of smart grids. The extensive installation of advanced metering infrastructure (AMI) brings convenience to consumers as well as utility providers by enabling real-time monitoring and management of energy consumption. However, dependence on communication technology also increases vulnerability to false data injection (FDI) or malicious attacks. In this paper, a model of machine learning known as an extreme gradient boosting classifier is explored by incorporating SCADA system for detection of anomalies in energy consumption patterns. Further, statistical tools such as principal component analysis and interquartile range are used for the preprocessing of the raw data. In this presented work, multi-class problem including six different types of synthetically developed attacks by considering statistical features are also examined. To validate the proposed framework, the IEEE 33 bus test system and the IRISH data set have been taken into consideration. Comparative analysis has been done with existing models using several evaluation matrices, such as accuracy, precision, recall, and F1 score. Results show that after integration of power system data, the accuracy of the proposed model is significantly improved.