Multivariate time-series cyberattack detection in the distributed secondary control of AC microgrids with convolutional neural network autoencoder ensemble

This paper proposes an unsupervised machine learning-based approach for cyberattack detection in AC microgrids with distributed secondary control architecture. The proposed approach is fully unsupervised and only utilizes the system’s normal datasets for the training of the algorithm. The attack und...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Sustainable Energy, Grids and Networks Ročník 38; s. 101374
Hlavní autoři: Roshanzadeh, Behshad, Choi, Jeewon, Bidram, Ali, Martínez-Ramón, Manel
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier Ltd 01.06.2024
Témata:
Convolutional neural networks
Cyberattack detection
Deep learning
Microgrid control
Deep learning
Convolutional neural networks
Microgrid control
Cyberattack detection
ISSN:2352-4677, 2352-4677
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:This paper proposes an unsupervised machine learning-based approach for cyberattack detection in AC microgrids with distributed secondary control architecture. The proposed approach is fully unsupervised and only utilizes the system’s normal datasets for the training of the algorithm. The attack under study is a false data injection (FDI) attack tampering with the operating frequency of inverter-based distributed generators (DGs). The paper utilizes a 1D Convolutional Autoencoder (CAE) for cyberattack detection on a microgrid’s distributed secondary frequency control. An autoencoder is a neural network architecture, where the model is trained to reconstruct its input in an unsupervised manner. CAE can be applied to a time-series dataset to extract features and exploit the known correlation between neighboring temporal features. Due to the correlation between the operating frequency of DGs and their active power ratios, the paper uses the time series of these two variables as inputs to CAE. The effectiveness of the proposed approach has been verified using a simulated microgrid test system in Matlab/Simulink.
ISSN:2352-4677
2352-4677
DOI:10.1016/j.segan.2024.101374