Sustainable Smart Energy Hub Design With V2H Interfaces: Protecting EV and Thermal Load Systems Under Disruption via Deep Q-Learning

As smart energy hubs (SEHs) evolve with the integration of electric vehicles (EVs) and their dynamic charging and discharging processes, they become increasingly susceptible to advanced cyberattacks, especially those targeting electrical and thermal demands. These attacks pose significant risks to integrated systems, including charging stations, vehicle-to-home (V2H) interfaces, and energy storage systems (ESS) comprising batteries for EVs and thermal storage (TS). The analysis focuses on how SEHs respond when subjected to stealthy data manipulation threats that evade detection, which exploit weaknesses in electrical and thermal loads to bypass conventional detection mechanisms and disrupt operational stability. To mitigate these threats, a robust machine learning (ML) framework based on reinforcement learning (RL) with deep Q-learning (DQN) is proposed. This approach reconstructs compromised load data, restoring system integrity and enhancing resilience. Test scenarios demonstrate that tampered inputs may cause electrical consumption to vary by nearly 17% and thermal usage by approximately 25%, severely impacting SEH stability. Following the deployment of the defense mechanism, the Mean Absolute Percentage Error (MAPE) is significantly reduced to 11.96% for electrical demand and 2.33% for thermal demand. The proposed RL-based strategy effectively enhances energy storage systems’ resilience against cyber threats, addressing key vulnerabilities and ensuring stable operations in smart grid ecosystems.