In the rapidly evolving landscape of power grid management, the integration of machine learning (ML) models has become a cornerstone for estimating excitation current in synchronous machines (SMs), a critical factor for maintaining grid stability and efficiency. However, this reliance on ML models has opened new avenues for cyber threats, particularly Denial of Service (DoS) attacks, which can disrupt grid operations and lead to significant financial and operational losses. A recent study published in the *Journal of Electrical Power and Energy Systems* sheds light on this pressing issue, offering insights into the vulnerabilities of ML models and proposing a robust defense strategy to mitigate the impact of such attacks.
Led by Najmul Alam from the Department of Electrical & Electronic Engineering at Rajshahi University of Engineering & Technology in Bangladesh, the research evaluates eight commonly used ML models—Gradient Boosting, Extreme Gradient Boosting (XGBoost), Categorical Boosting (CatBoost), Random Forest, Long Short-Term Memory (LSTM), Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), and ridge regression—in estimating excitation current under normal, attacked, and defense scenarios. The study employs key metrics such as Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R2) to assess model performance.
The findings reveal that tree-based models, particularly Gradient Boosting and Random Forest, deliver superior accuracy under normal conditions. However, under DoS attack conditions, these models exhibit significant performance degradation, with RMSE increasing by up to 6295%. This stark decline in performance underscores the vulnerability of ML models to cyber threats, highlighting the urgent need for robust defense mechanisms.
To address this challenge, Alam and his team proposed a defense strategy that incorporates a detection system to identify and replace attacked data with predicted values. This approach not only restores model performance but also enhances the reliability of estimation systems under attack conditions, reducing RMSE by up to 2725%. “Our defense strategy effectively mitigates the impact of DoS attacks, ensuring the accuracy and reliability of excitation current estimation in power grids,” Alam explained. “This is crucial for maintaining grid stability and efficiency, especially in the face of increasing cyber threats.”
The implications of this research extend beyond the immediate context of power grid management. As the energy sector continues to embrace digital transformation, the integration of cyber-resilient ML models will be pivotal in safeguarding critical infrastructure against cyber threats. The proposed defense strategy offers a practical solution for enhancing the resilience of ML applications in power grid systems, paving the way for more secure and efficient energy management practices.
“This study contributes to the development of robust and cyber-resilient estimating methodologies, offering practical implications for the advancement of ML applications in power grid systems and beyond,” Alam noted. The research not only highlights the vulnerabilities of ML models but also provides a roadmap for building more secure and reliable energy systems, ensuring the stability and efficiency of power grids in an increasingly digital world.
As the energy sector navigates the complexities of cyber threats, the insights from this study will be invaluable for energy professionals, policymakers, and researchers. By adopting robust defense strategies, the energy sector can mitigate the impact of DoS attacks and ensure the seamless operation of power grids, ultimately contributing to a more secure and sustainable energy future.