Machine Learning-assisted Energy Management System for an Islanded Microgrid and Investigation of Data Integrity Attack on Power Generation

This paper presents an integrated energy management system (EMS) for an islanded microgrid, and briefly investigates the system’s performance in case of false load injection attack. The proposed energy management system includes a Multi-step Deep LSTM neural network, and a mixed-integer optimization algorithm. The Deep LSTM neural network forecasts the load data, while the optimizer determines the best setpoints for the microgrid’s decentralized controllers. The EMS is integrated with an islanded microgrid to evaluate the viability of the proposed system. It is also shown that Cyber-physical attack causes frequency deviation in the microgrid system.