Fakhrooeian Mahan, Basem Ali, Gholami Mohammad Mahdi, Iliaee Nahal, Amidi Alireza Mohammadi, Hamzehkanloo Amin Heydarian, Karimipouya Akbar
Institute for Electrical Machines, Traction and Drives, Technische Universität Braunschweig, 38106, Braunschweig, Germany.
Air Conditioning Engineering Department, College of Engineering, University of Warith Al-Anbiyaa, Karbala, Iraq.
Sci Rep. 2025 Jan 30;15(1):3762. doi: 10.1038/s41598-025-88090-5.
Microgrid systems have evolved based on renewable energies including wind, solar, and hydrogen to make the satisfaction of loads far from the main grid more flexible and controllable using both island- and grid-connected modes. Albeit microgrids can gain beneficial results in cost and energy schedules once operating in grid-connected mode, such systems are vulnerable to malicious attacks from the viewpoint of cybersecurity. With this in mind, this paper explores a novel advanced attack model named the false transferred data injection (FTDI) attack aiming to manipulatively alter the power flowing from the microgrid to the upstream grid to raise voltage usability probability. One crucial piece of information that the model uses to change the system and cause the greatest amount of damage while concealing the attacker's view is the voltage stability index. Saying that the power transaction between the microgrid and the upstream grid is within the broad scope of bilateral exchange at any given moment is noteworthy. Put otherwise, with respect to the FTDI assault, the microgrid's power direction is just as significant to the detection system as the transferred power value. Therefore, once the microgrid is running in the grid-connected mode, the false data detector needs to concurrently detect changes in the value and direction of power. To overcome this problem, the paper presents a learning generative network model, based on the generative adversarial network (GAN) paradigm, to recognize the change in probability values that is maliciously aimed. To this end, a studied microgrid system including the wind turbine, photovoltaic, storage, tidal turbine, and fuel cell units is performed on the tested 24-bus IEEE grid to satisfy the local load demands. Comparative analysis indicates notable gains, such as scores of 0.95%, 0.92%, 0.7%, and 10% for the Hit rate, C.R. rate, F.A. rate, and Miss rate in order to evaluate the GAN-based detection model within the microgrid.
微电网系统已基于包括风能、太阳能和氢能在内的可再生能源发展而来，以便通过孤岛运行模式和并网运行模式，更灵活、可控地满足远离主电网的负荷需求。尽管微电网在并网运行模式下，在成本和能源调度方面能取得有益成果，但从网络安全角度来看，此类系统容易受到恶意攻击。考虑到这一点，本文探索了一种名为虚假传输数据注入（FTDI）攻击的新型高级攻击模型，旨在通过操纵改变从微电网流向上游电网的功率，以提高电压可用性概率。该模型用于改变系统并在隐藏攻击者意图的同时造成最大程度破坏的一个关键信息是电压稳定性指标。值得注意的是，在任何给定时刻，微电网与上游电网之间的电力交易都处于双边交换的广泛范围内。换句话说，对于FTDI攻击而言，微电网的功率方向对检测系统与传输功率值同样重要。因此，一旦微电网以并网模式运行，虚假数据检测器需要同时检测功率值和方向的变化。为克服这一问题，本文提出了一种基于生成对抗网络（GAN）范式的学习生成网络模型，以识别恶意攻击的概率值变化。为此，在经过测试的24节点IEEE电网中运行一个经过研究的微电网系统，该系统包括风力涡轮机、光伏、储能、潮汐涡轮机和燃料电池单元，以满足当地负荷需求。对比分析表明，为评估微电网内基于GAN的检测模型，在命中率、误报率、拒识率和漏报率方面取得了显著成效，分别为0.95%、0.92%、0.7%和10%。
