Robust Power System Stability Assessment Against Adversarial Machine Learning-Based Cyberattacks via Online Purification

Zhao, Tianqiao; Yue, Meng; Wang, Jianhui

doi:10.1109/tpwrs.2022.3233735

Cited by 5 publications

(14 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhang et al, 2023aR. Zhang et al, , 2023bZhao et al, 2023aZhao et al, , 2023bZhu, Hill, & Lu, 2021;Zhu & Luo, 2021). The last AI technique found in this review is Quantum Machine Learning (QML) which implements a Variational Quantum Circuit (an ordered sequence of quantum gates characterized by variable parameters that can be tuned during the training procedure) that takes advantage of the ability of quantum machines to perform complex tensor operations and utilize quantum operators (such us superposition and entanglement) to capture complex non-linear behavior is the first (and only work so far) that implements QML for DSA.…”

Section: Types Of Techniquementioning

confidence: 99%

“…Zhang et al, , 2023bY. Zhang, Dong, et al, 2018;Zhao et al, 2023b;Zhu et al, 2017)). All these methods were proposed to detect fast (online) instabilities after a contingency, where prediction speed is crucial.…”

Section: Voltage Stabilitymentioning

confidence: 99%

“…Li et al, 2021;C. Wang et al, 2021;Yousefian et al, 2017;Yousefian & Kamalasadan, 2016Zhao et al, 2023a;Zhou et al, 2016), the 74-bus Nordic System (Naderi et al, 2023), the 77-bus Nordic System , the simplified Australian Electric System of 59 buses and 14 generators (R. Liu et al, 2018;Suprême et al, 2018), the Indian 75-bus system (Chang et al, 2005;Jain et al, 2003), the Iranian 66-bus power grid (H. Mohammadi et al, 2018Mohammadi et al, , 2016, the IEEE standard 118-bus system (Anderson et al, 2023;Dharmapala et al, 2020;Maihemuti et al, 2021;Mukherjee & De, 2020;Nie et al, 2017;Xie & Sun, 2021;Zhao et al, 2023b), the 123-bus IEEE system (Y. Wang & Pal, 2023), the 145-bus IEEE 50-generator test system (Azman et al, 2020;Velayati et al, 2015;Y.…”

Section: Test Systemsmentioning

confidence: 99%

“…Zhang et al, , 2015, the 162-bus IEEE 17 machine (Yu et al, 2018), the 300-bus IEEE system , the 179-bus WECC system (Mahdi & Genc, 2018;Malbasa et al, 2017;Senroy et al, 2006;Zheng et al, 2013), the 127-bus WSCC (Mahdi & Genc, 2017), the 181-bus Indian system (Suganyadevi et al, 2016), the 120-bus dynamic equivalent system of Chinas power grid load (Y. Xu et al, 2012), the 176-bus Pacific AC Intertie (PACI) (Rovnyak et al, 2000), the 140-bus NPCC (Q. , the 200-bus Illinois system (Ren, Wang, et al, 2023;Ren, Yuan, et al, 2023;Zhao et al, 2023b), the 360-bus subsystem within AEP (Nuqui et al, 2001), and the 153-bus North GZ Power Grid in China (Zhu & Luo, 2021). Among the large systems used, there is only the 1648-bus PSS/E (S. .…”

Section: Test Systemsmentioning

confidence: 99%

See 3 more Smart Citations

Artificial Intelligence Techniques for Dynamic Security Assessments – A Survey

Cuevas,

Alvarez-Malebrán,

Rahmann

et al. 2024

Preprint

View full text Add to dashboard Cite

The rise of converter interface generation (CIGs) technologies with fast response times has introduced new challenges for the stability of electrical power systems. Traditional dynamic security assessment (DSA) methods of power systems are based on dynamic simulations within a time domain. Because of the high computational efforts and human resources required to carry out these kinds of simulations, they are performed for a limited number of adverse (critical) operational conditions in terms of their stability. However, further incorporation of CIGs in power systems not only results in a shift of the critical operating conditions that may threaten the system’s stability, but also the number of critical conditions increases. Consequently, there is an urgent need to develop new methods for evaluating system stability in a wide range of operating conditions with a reasonable amount of human and computational effort. A complementary 1 strategy to the traditional methods for DSA is to use artificial intelligence (AI) techniques. Unlike traditional methods, AI techniques use a data set that captures the non-linear relationships between the system’s operational conditions and their stability, without needing to solve the algebraic-differential equations modeling the power system. Once these relationships are established, the system’s stability can be evaluated for other operational conditions much faster (within the order of the hundreds of milliseconds) and accurately. This, in turn, allows for the consideration of an extended range of operational conditions for DSA studies. This article surveys the state of the art regarding the use of AI techniques for DSA in electrical power systems. Focus is made on algorithms, the kind of stability being addressed, data processing and applications. Finally, limitations, challenges, and future trends are discussed.

show abstract

Section: Types Of Techniquementioning

confidence: 99%

Section: Voltage Stabilitymentioning

confidence: 99%

Section: Test Systemsmentioning

confidence: 99%

Section: Test Systemsmentioning

confidence: 99%

See 2 more Smart Citations

Artificial Intelligence Techniques for Dynamic Security Assessments – A Survey

Cuevas,

Alvarez-Malebrán,

Rahmann

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…The susceptibility to adversarial examples undermines the ability of ML models to make accurate and reliable predictions, potentially leading to severe consequences. In the case of PQDs, the misclassification of such events could result in incorrect decision-making and inadequate response measures, jeopardizing the stability of the power grid system and the safety of connected devices [12], [13].…”

Section: Introductionmentioning

confidence: 99%

Mitigating Targeted Universal Adversarial Attacks on Time Series Power Quality Disturbances Models

Khan,

Mynuddin,

Adom

et al. 2023

2023 5th IEEE International Conference on Trust, Privacy and Security in Intelligent Systems and Applications (TPS-ISA)

View full text Add to dashboard Cite

The utilization of deep learning models has been widely recognized for its significant contribution to the enhancement of smart grid operations, particularly in the domain of power quality disturbance (PQD) classification. Nevertheless, the emergence of vulnerabilities like targeted universal adversarial attacks can significantly undermine the reliability and security of deep learning models. These attacks can exploit the model's weaknesses, causing it to misclassify PQDs with potentially catastrophic consequences. In our previous research, we for the first time examined the vulnerability of deep learning models to targeted universal adversarial attacks on time series data in smart grids by introducing a novel algorithm that effectively attacks by maintaining a trade-off between fooling rate and imperceptibility. While this attack method demonstrated notable efficacy, it also emphasized the pressing need for robust defensive mechanisms to safeguard these critical systems. This paper provides a thorough examination and evaluation of different defense strategies, specifically adversarial training, defensive distillation, and feature squeezing, in order to identify the most effective method for mitigating targeted universal adversarial (TUA) attacks on time series data for three different types of imperceptibility (high, medium and low). Based on our analysis, adversarial training demonstrates a significant reduction in the success rate of attacks. Specifically, the technique reduced fooling rates by an average of 23.73% for high imperceptibility, 31.04% for medium imperceptibility, and a substantial 42.96% for low imperceptibility. These findings highlight the crucial role of adversarial training in enhancing the integrity of deep learning applications.

show abstract