Stochastic Energy Management and Scheduling of Microgrids in Correlated Environment: A Deep Learning-Oriented approach

Abstract Regarding the operation, reliability, security, and cost-effectiveness of microgrids (MGs), optimal energy management and data security are essential issues that must be taken into consideration. In this regard, this paper proposes a secured stochastic energy management scheme for both grid-connected and islanded hybrid AC-DC MGs (HMGs) considering renewable energy sources (RESs), plug-in hybrid electric vehicles (PHEVs), and energy storage devices (ESs). In this work, the operation of HMGs is formulated as a single objective optimization problem, which is solved using the interior search optimization algorithm (ISOA). In order to reinforce data security in HMGs, a deep learning-based intrusion detection system (IDS) based on the long short-term memory (LSTM) and prediction interval is proposed to detect false data injection attacks (FDIA) on advanced metering infrastructures (AMIs). In this work, the 2 m point estimation method is utilized to model uncertainties associated with the forecasted output power of RESs, energy market price, charging demand of PHEVs, and load demand. The performance of the proposed scheme is evaluated using IEEE 33 bus test system for both operation modes.