Detection of False Data Injection Cyberattacks Targeting Smart Transmission/Distribution Networks

Abstract

Integration of different infrastructures in the normal operation of modern-day power systems paves the way for adversaries to penetrate into the system and manipulate the data. This is because modern power grids need to be monitored/controlled via information and communication technology (ICT). One of the most significant challenges associated with such power networks is the risk of operational issues (e.g., congestion, voltage instability, etc.) as a consequence of stealthy false data injection (FDI) cyberattacks bypassing bad data detection (BDD) algorithms embedded in both DC and AC state estimations. Toward this end, this paper develops a detection framework oriented toward multi-layer perceptron (MLP) networks to protect the measurements to be processed by power system operators in the upper level. Levenberg-Marquardt (LM) backpropagation, which is a network training function updating the weight/bias based on LM optimization, is implemented to train the developed neural network (NN). A mean absolute percentage error and mean of squares of the network errors are considered to assess the accuracy of the prediction. The developed feed-forward neural network tracks the measurements (e.g., active and reactive powers, voltage magnitudes, etc.) to find the relationship between them to warn the system operator in case of systematic manipulation in the dataset. The effectiveness of the proposed detection framework is validated on the IEEE 14-bus transmission system and the IEEE 33-bus distribution grid.