Machine learning based false data injection in smart grid

Abstract

Smart Grid is the seamless integration of advance digital communication network, state of the art control technologies, and power system infrastructure working together as an entity to ensure the reliability, sustainability, and stability of the power infrastructure. Digital communication network with is the key to the reliability of Smart Grid as all control actions are deemed upon the data transmitted by a communication network. With false data, however, the same digital communication network can lead to anomalies like abnormal disruptions, load shedding, malicious attacks and power theft. Robust False data injection attack methods proposed till now demand for the complete knowledge of interconnected power grid network topology. In this paper, three network topology independent techniques for false data injection into the smart grid are proposed based on linear regression, linear regression with time stamp, and by using delta thresholds. To make injected false data more unlikely to be detected, it is constructed to fill up the missing measurements in real-time data. The robustness of proposed attack algorithms are stated by state-of-the-art defence techniques, i.e. Bad Data Detection, AC State estimation, Support Vector Machine, and Temporal Behaviours based False data detection.