LSTM Networks for Cyber-physical Attack Diagnoses in Microgrids.

This work is a follow-up to the experimentation that used the NARX neural networks as an intelligent detecting mechanism to detect cyber-physical attacks in connected microgrids. In this paper, we continue to explore the practical dimensions and limitations of the NRAX. Simultaneously, we propose the LSTM model to circumvent the encountered constraints. The trained LSTM model is then tested and proven effective in detecting attacks over a selection of both cyber and physically introduced attacks.