Resilient Frequency Regulation for DoS Attack Intensity Adaptation via Predictive Reinforcement V2G Control Learning
Jian Sun, Xin Wang, Guanqiu Qi, Huaqing Li, Xin Wang, Huiwei Wang, Juan C. Vásquez, Josep M. Guerrero
Abstract
Large-scale integration of renewable energy sources (RES) into power grids brings unpredictable intermittent power generation, leading to power grid frequency excursions. Electric vehicles (EVs) using vehicle-to-grid (V2G) technology are responsive and cost-effective, providing an effective alternative to power grid frequency regulation (FR). However, EVs inevitably use commonly shared communication networks, which are susceptible to denial-of-service (DoS) attacks, significantly degrading V2G-based FR (V2G-FR) performance. To optimize V2G-FR systems under DoS attacks, this paper proposes a multi-step predictive reinforcement learning V2G control (MPRLC) scheme. The FR performance degradation is mitigated by predicting multiple control steps blocked by DoS attacks. A reinforcement learning (RL) framework is built to achieve predictions without the need for a system model, enabling the V2G-FR controller to adapt to changes in the power system. In addition, the number of transmitted predictive control steps (NTPCS) is proposed to adapt to time-varying attack intensity, thereby further improving control performance. The effectiveness and advantages of the MPRLC have been verified on the IEEE 39-bus system. The results show that the MPRLC can effectively compensate for control signals interfered by attacks. The results also indicate that the NTPCS should increase to provide adequate compensation as attack intensity increases.