Artificial Neural Network and Peer-to-Peer Communications at the Grid-Edge to Mitigate Cyber Attacks on Distributed Photovoltaic Inverters

Resilient control of photovoltaic (PV) inverters using a local Artificial Neural Network (ANN) and peer-to-peer communications can maintain grid services during a cyberattack. High penetrations of PV systems presents grid performance challenges, and alterations to connected systems can introduce additional problems. To tackle these issues, this paper introduces a methodology for controlling PV inverters that are under attack using the Laterally Primed Adaptive Resonance Theory (LAPART) ANN to predict the best reactive power control input when communications between the central command center are down or cannot be trusted. This work tested the approach using a 6-bus feeder model with a high penetration of PV. The experiment found that the algorithm can predict the appropriate reactive power setting with high accuracy, and when embedded inside the grid model, the algorithm can predict a reactive power that improved system voltages.