A data-driven resilience framework for renewable telecom microgrids under cyber–physical attacks

Renewable energy-driven IoT microgrids, such as telecommunication base stations, are increasingly vulnerable to sophisticated cyber–physical attacks due to their deep digitalization. Traditional rule-based intrusion detection systems struggle to differentiate between natural physical non-linearities and stealthy threats, such as sensor data spoofing (Chameleon attacks) and control logic tampering (Vampire attacks), and often lack actionable recovery mechanisms. To address this critical gap, this paper proposes a Resilience-Oriented Control Framework (ROCF) based on a hierarchical intelligent control architecture. The framework decouples the defense mechanism into an AI-driven perception layer, utilizing sequence-based deep learning to identify cross-modal semantic contradictions, and a rule-based execution layer that triggers deterministic mitigation protocols. We evaluate the framework using a high-fidelity digital twin under a parallel simulation setup across diverse attack scenarios and stochastic environmental seeds. Experimental results demonstrate that the proposed LSTM-based perception layer achieves an outstanding F1-score of 0.9804 with minimal variance. Upon detection, the ROCF actively mitigates the threats: it reduces the battery stress index by nearly 80% during logic tampering and cuts unnecessary grid purchase costs by up to 49% during sensor spoofing, while leaving the protected normal-operation KPI trajectory unchanged. By combining the pattern-recognition power of AI with the explainability of physical rules, the proposed framework provides a robust and highly deployable solution for enhancing the cyber–physical resilience of next-generation smart energy infrastructures.