ABSTRACT
This chapter considers the smart grid security problem from a defender’s perspective and provides an effective detection scheme using reinforcement learning (RL) techniques. It presents an online cyber-attack detection algorithm using the framework of model-free RL for partially observable Markov decision process. There are many types of cyber-attacks, among them false data injection, jamming, and denial of service attacks are well known. The main objective of attackers is to damage or mislead the state estimation mechanism in the smart grid to cause wide-area power blackouts or to manipulate electricity market prices. Since the smart grid is regulated based on estimated system states, state estimation is a fundamental task in the smart grid, that is conventionally performed using the static least square estimators. For a discrete-time linear dynamic system, if the noise terms are Gaussian, the Kalman filter is the optimal linear estimator in minimizing the mean squared state estimation error.
