Risk Assessment of Distributed PV Microgrids Under Cyber Attacks based on CNN-AE

The widespread application of smart grid technology and numerous Internet of Things (IoT) devices expose microgrid vulnerability to various cyber attacks. Moreover, the complex interactions and implicit risk propagation within the system pose challenges in accessing the risk from cyber attacks. To this end, we propose a risk assessment method for distributed PV microgrids under cyber attacks. This program employs an autoencoder (AE) architecture where convolutional neural networks (CNN) serve as the intermediate layer to extract the local features of data exchange from adjacent nodes. The AE conducts unsupervised learning using power data under normal operating conditions, thus performing risk assessment by evaluating the reconstruction errors. To validate the proposed method, we conducted training on the algorithm using historical data and the topology of a real scenario of a distributed photovoltaic microgrid and tested the risk assessment with various fake data injection attacks (FDIA) and Denial of Service (DoS) attacks.