Sarcouncil Journal of Engineering and Computer Sciences

Abstract: The incorporation of decentralized solar microgrids in modern electrical grids has promoted energy efficiency and sustainability, but such addition constructs a broad common attack surface in interconnected infrastructures. Traditional Intrusion Detection Systems (IDSs) may find it difficult to accommodate the dynamic and data-rich scenario in smart microgrids; hence, there is demand for predictive and intelligent cybersecurity approaches. One such solution is characterized in this work, which investigates the use of machine learning (ML) for predicting cyber threats in solar microgrids and preventing attacks, including false data injection, denial-of-service and spoofing. A ML hybrid model(integrated with CNN and LSTM networks) was proposed to extract the spatial-temporal feature from sensor data and communication network flow. Test results show that the proposed model can reach 96.8% detection accuracy and it has a lower false alarm rate than rule based method does. The combination of ML-based predictive analytics offers up-to-the-minute situational awareness for proactive threat mitigation and operational resiliency. This study demonstrates the promise of smart data-driven solutions for improving cybersecurity in renewable energy networks and contributes to the efforts towards self-learning defense mechanisms for future solar microgrids.