Similar articles

Abstract: load frequency control (LFC) is usually managed by traditional proportional–integral–derivative (PID) controllers. Recently, deep reinforcement learning (DRL)-based adaptive controllers have been widely studied for their superior performance. However, the DRL-based adaptive controller exhibits inherent vulnerability due to adversarial attacks. To develop more robust control systems, this study conducts a deep analysis of DRL-based adaptive controller vulnerability under adversarial attacks. First, an adaptive controller is developed based on the DRL algorithm. Subsequently, considering the limited capability of attackers, the DRL-based LFC is evaluated under adversarial attacks using the zeroth-order optimization (ZOO) method. Finally, we use adversarial training to enhance the robustness of DRL-based adaptive controllers. Extensive simulations are conducted to evaluate the performance of the DRL-based PID controller with and without adversarial attacks.

面向负载频率控制场景下深度强化学习比例-积分-微分控制器的黑盒对抗攻击

[1]Albeladi F, Barati M, 2023. Grid-supportive load frequency control using deep reinforcement learning. Proc IEEE Kansas Power and Energy Conf, p.1-5.

[2]Behzadan V, Munir A, 2017. Vulnerability of deep reinforcement learning to policy induction attacks. Proc 13^th Int Conf on Machine Learning and Data Mining in Pattern Recognition, p.262-275.

[3]Chaojun G, Jirutitijaroen P, Motani M, 2015. Detecting false data injection attacks in AC state estimation. IEEE Trans Smart Grid, 6(5):2476-2483.

[4]Chen PY, Zhang H, Sharma Y, et al., 2017. ZOO: zeroth order optimization based black-box attacks to deep neural networks without training substitute models. Proc 10^th ACM Workshop on Artificial Intelligence and Security, p.15-26.

[5]Chen ST, Liu GJ, Zhou ZY, et al., 2024. Robust multi-agent reinforcement learning method based on adversarial domain randomization for real-world dual-UAV cooperation. IEEE Trans Intell Veh, 9(1):1615-1627.

[6]Chen TM, 2010. Stuxnet, the real start of cyber warfare? IEEE Netw, 24(6):2-3.

[7]Doan DV, Nguyen K, Thai QV, 2022. Load-frequency control of three-area interconnected power systems with renewable energy sources using novel PSO PID-like fuzzy logic controllers. Eng Technol Appl Sci Res, 12(3):8597-8604.

[8]Dogru O, Velswamy K, Ibrahim F, et al., 2022. Reinforcement learning approach to autonomous PID tuning. Comput Chem Eng, 161:107760.

[9]Fujimoto S, Hoof H, Meger D, 2018. Addressing function approximation error in actor-critic methods. Proc 35^th Int Conf on Machine Learning, p.1582-1591.

[10]Gleave A, Dennis M, Wild C, et al., 2020. Adversarial policies: attacking deep reinforcement learning. Proc 8^th Int Conf on Learning Representations.

[11]Guo WR, Liu GJ, Zhou ZY, et al., 2024. Enhancing the robustness of QMIX against state-adversarial attacks. Neurocomputing, 572:127191.

[12]Hao JB, Tao Y, 2022. Adversarial attacks on deep learning models in smart grids. Energy Rep, 8:123-129.

[13]Jia XJ, Zhang Y, Wu BY, et al., 2022. LAS-AT: adversarial training with learnable attack strategy. Proc IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.13388-13398.

[14]Liu XH, Jiao QM, Yan ZM, 2023. Load frequency control with deep reinforcement learning under adversarial attacks. Proc 18^th Conf on Industrial Electronics and Applications, p.257-262.

[15]Madry A, Makelov A, Schmidt L, et al., 2018. Towards deep learning models resistant to adversarial attacks. Proc 6^th Int Conf on Learning Representations.

[16]Maei HR, 2011. Gradient Temporal-Difference Learning Algorithms. PhD Thesis, Department of Computing Science, University of Alberta, Canada.

[17]Michel A, Jha SK, Ewetz R, 2022. A survey on the vulnerability of deep neural networks against adversarial attacks. Prog Artif Intell, 11(2):131-141.

[18]Mishchenko D, Oleinikova I, Erdődi L, et al., 2024. Multidomain cyber-physical testbed for power system vulnerability assessment. IEEE Access, 12:38135-38149.

[19]Moldovan D, Ayyanar R, 2024. DNP3 implementation in a high DER penetration distribution system. Proc IEEE Kansas Power and Energy Conf, p.1-5.

[20]Muduli R, Jena D, Moger T, 2025. Application of reinforcement learning-based adaptive PID controller for automatic generation control of multi-area power system. IEEE Trans Automat Sci Eng, 22:1057-1068.

[21]Muhammad MS, Alshra‘a AS, German R, 2024. Survey of cybersecurity in smart grids protocols and datasets. Proc Comput Sci, 241:365-372.

[22]Nafees MN, Saxena N, Cardenas A, et al., 2023. Smart grid cyber-physical situational awareness of complex operational technology attacks: a review. ACM Comput Surv, 55(10):215.

[23]Pandey SK, Gupta P, Dwivedi SS, 2020. Full order observer based load frequency control of single area power system. Proc 12^th Int Conf on Computational Intelligence and Communication Networks, p.239-242.

[24]Pattanaik A, Tang ZY, Liu SJ, et al., 2018. Robust deep reinforcement learning with adversarial attacks. Proc 17^th Int Conf on Autonomous Agents and Multiagent Systems, p.2040-2042.

[25]Qassim QS, Ali MAM, Tahir NM, 2023. Security analysis of DNP3 protocol in SCADA system. Proc 13^th Int Conf on Control System, Computing and Engineering, p.314-319.

[26]Qiaoben Y, Ying CY, Zhou XN, et al., 2024. Understanding adversarial attacks on observations in deep reinforcement learning. Sci China Inform Sci, 67(5):152104.

[27]Raju GV, Srikanth NV, 2024. Frequency control of an islanded microgrid with multi-stage PID control approach using moth-flame optimization algorithm. Electr Eng, 107:8861-8878.

[28]Rasolomampionona DD, Połecki M, Zagrajek K, et al., 2024. A comprehensive review of load frequency control technologies. Energies, 17(12):2915.

[29]Saxena S, Bhatia S, Gupta R, 2021. Cybersecurity analysis of load frequency control in power systems: a survey. Designs, 5(3):52.

[30]Shabani H, Vahidi B, Ebrahimpour M, 2013. A robust PID controller based on imperialist competitive algorithm for load-frequency control of power systems. ISA Trans, 52(1):88-95.

[31]Shangguan XC, Zhang CK, He Y, et al., 2021. Robust load frequency control for power system considering transmission delay and sampling period. IEEE Trans Ind Inform, 17(8):5292-5303.

[32]Shi HR, Liu GJ, Zhang KW, et al., 2023. MARL sim2real transfer: merging physical reality with digital virtuality in metaverse. IEEE Trans Syst Man Cybern Syst, 53(4):2107-2117.

[33]Shuprajhaa T, Sujit SK, Srinivasan K, 2022. Reinforcement learning based adaptive PID controller design for control of linear/nonlinear unstable processes. Appl Soft Comput, 128:109450.

[34]Takiddin A, Ismail M, Serpedin E, 2023. Robust data-driven detection of electricity theft adversarial evasion attacks in smart grids. IEEE Trans Smart Grid, 14(1):663-676.

[35]Tan KL, Esfandiari Y, Lee XY, et al., 2020. Robustifying reinforcement learning agents via action space adversarial training. Proc American Control Conf, p.3959-3964.

[36]Tian JW, Wang BH, Li J, et al., 2022. Adversarial attacks and defense for CNN based power quality recognition in smart grid. IEEE Trans Netw Sci Eng, 9(2):807-819.

[37]Xue J, Liu ZN, Liu GJ, et al., 2024. Robust wind-resistant hovering control of quadrotor UAVs using deep reinforcement learning. IEEE Trans Intell Veh, early access.

[38]Yan S, Gu ZH, Park JH, 2021. Memory-event-triggered H_∞load frequency control of multi-area power systems with cyber-attacks and communication delays. IEEE Trans Netw Sci Eng, 8(2):1571-1583.

[39]Yan ZM, Xu Y, 2019. Data-driven load frequency control for stochastic power systems: a deep reinforcement learning method with continuous action search. IEEE Trans Power Syst, 34(2):1653-1656.

[40]Zhang LH, Jiang CM, Pang AP, 2022. Black-box attacks and defense for DNN-based power quality classification in smart grid. Energy Rep, 8:12203-12214.

[41]Zhang ZY, Deng RL, Yau DKY, et al., 2021. Zero-parameter-information data integrity attacks and countermeasures in IoT-based smart grid. IEEE Int Things J, 8(8): 6608-6623.

[42]Zhang ZY, Yang ZB, Yau DKY, et al., 2023a. Data security of machine learning applied in low-carbon smart grid: a formal model for the physics-constrained robustness. Appl Energy, 347:121405.

[43]Zhang ZY, Deng RL, Tian YL, et al., 2023b. SPMA: stealthy physics-manipulated attack and countermeasures in cyber-physical smart grid. IEEE Trans Inform Forensics Secur, 18:581-596.

[44]Zhang ZY, Yang KD, Tian YL, et al., 2024a. An anti-disguise authentication system using the first impression of avatar in metaverse. IEEE Trans Inform Forensics Secur, 19:6393-6408.

[45]Zhang ZY, Deng RL, Yau DK, 2024b. Vulnerability of the load frequency control against the network parameter attack. IEEE Trans Smart Grid, 15(1):921-933.

[46]Zhang ZY, Liu MX, Sun MY, et al., 2024c. Vulnerability of machine learning approaches applied in IoT-based smart grid: a review. IEEE Int Things J, 11(11):18951-18975.

[47]Zheng Y, Yan ZM, Chen KJ, et al., 2021. Vulnerability assessment of deep reinforcement learning models for power system topology optimization. IEEE Trans Smart Grid, 12(4):3613-3623.

[48]Zhou ZY, Liu GJ, Guo WR, et al., 2024a. Adversarial attacks on multiagent deep reinforcement learning models in continuous action space. IEEE Trans Syst Man Cybern Syst, 54(12):7633-7646.

[49]Zhou ZY, Liu GJ, Zhou MC, 2024b. A robust mean-field actor-critic reinforcement learning against adversarial perturbations on agent states. IEEE Trans Neur Netw Learn Syst, 35(10):14370-14381.