Blind Topology Identification for Smart Grid Based on Dictionary Learning

Abstract

Malicious attacks, especially false data injection attacks, pose a huge threat to the safe and stable operation of smart grid. In order to better understand the attack process and propose a defense scheme, this paper identifies the smart grid structure blindly from the perspective of attacker. In detail, we first collect the injected active power data of each bus, then a dictionary learning-based method is proposed to solve the blind topology identification task. In order to enhance the accuracy of the topology identification problem, four underlying features of smart grid structure are extracted by the investigation of power network topology. Lastly, we incorporate these features into the framework of dictionary learning as prior knowledge so as to improve the identification performance. Through intensive experiments on the IEEE-14 bus system, it can be found that the proposed method can well identify the location of network edges. Furthermore, we can find that the identified smart grid structure retains some important graph properties with the original smart grid.