부산대학교 도서관

The smart integration of communication and information technology into electrical power grids has significantly benefited the monitoring and operations of the grid, but it has also made it vulnerable to cyber attacks. A Load Redistribution (LR) attack is a type of attack in which the attacker manipulates one or more load requests to create an overflow (congestion) in a transmission line. The cost and limitations of conventional detectors have motivated researchers to explore Machine Learning (ML) and data driven mechanisms for detecting potential LR attacks. The expanding nature of the grid and the rapidly increasing number of loads on a grid necessitates the monitoring of a huge number of points and creates a data deluge for the ML-based detectors, thereby adversely affecting their performance. This paper provides an approach for identifying a subset of loads, namely critical loads, that needs to be monitored for detecting LR attacks. This helps the grid operator to monitor fewer loads thereby reducing the cost of monitoring the loads, the detection time, and the computational burden. The proposed approach is implemented on the IEEE 30 bus system. The results validate the fact that LR attacks can be detected by monitoring only the identified critical loads, without adversely affecting the performance of the ML detectors.