Attack Detection in Automatic Generation Control Systems using LSTM-Based Stacked Autoencoders

Abstract

Automatic generation control (AGC) is paramount in maintaining the stability and operation of power grids. Its dependence on communication systems makes it vulnerable to various cyberphysical attacks. False data injection attacks (FDIA) are particularly difficult to detect and represent a major threat to AGC systems. This article proposes a novel spatio-temporal learning algorithm that can learn the normal dynamics of the power grid with AGC system to deal with this problem. The algorithm first uses a long short-term memory autoencoder to learn the normal dynamics. It then utilizes this unsupervised learned model in detecting the various possibilities of FDIA affecting the AGC system by evaluating the reconstruction residual of each measurements sample. The proposed algorithm is data-driven which makes it resilient against AGC's parameters uncertainties and modeling nonlinearities. The effectiveness of the developed algorithm is evaluated through test cases with various basic and stealth FDIAs.