Cyber attacks detection using deep learning methods for resilient operation in DC shipboard microgrids

Abstract

Cyber resilience has become paramount as a transition of maritime systems towards digitization, particularly within DC shipboard microgrids (SMGs). Adopting innovative communication technologies can enhance the resilience of SMGs for stable operation. However, challenges like false data injection and Man-in-The-Middle attacks pose significant threats to SMG operations when integrating these technologies into ship intelligent grids. In this regard, this paper proposes a reliable deep learning (DL) method, especially an Artificial neural network (ANN) with a deep encoder-decoder architecture, for the detection of cyber-intrusions and mitigating their effects, ensuring system control and stability for resilient operation of SMG. Detecting malicious data intrusions is crucial for maintaining optimal grid conditions and preventing disruptions in load dispatch. The proposed method utilizes a fusion of current and voltage data features for comprehensive DL model training, resulting in an adequate level of detection accuracy and providing cybersecurity analysis for SMGs, addressing component-level attacks, and devising defense strategies from aspects of detection, mitigation, and prevention. Furthermore, the deep ANN is fine-tuned with optimal hyperparameters to effectively counter cyberattacks, achieving an enhanced accuracy rate of 97.51% and minimal loss of 0.101%, surpassing conventional machine learning approaches. Rigorous test scenarios are performed to validate the robustness of the proposed method, emphasizing the cyber resilience of DC SMGs for enhanced security and operational integrity.