Chenjiao Bai∗
Digital twin, power equipment inspection, deep learning, knowledge distillation
This paper presents a digital twin-based real-time inspection and monitoring method for power equipment, combining lightweight deep learning with dynamic virtual modeling to achieve high- precision fault diagnosis and predictive maintenance. The proposed framework integrates a four-layer architecture (physical sensing, virtual simulation, intelligent analysis, and closed-loop control) to enable accurate state mirroring and autonomous decision-making. By enhancing ShuffleNetV2 with Lightweight SimAM attention and knowledge distillation, the model achieves 97.5% accuracy while maintaining low computational overhead, significantly outperforming conventional approaches. Experimental results validate the method’s superior balance between efficiency and accuracy and its practical applicability in resource-constrained scenarios. This study advances intelligent power equipment management by unifying digital twin simulation with edge-deployable AI, offering a scalable solution for modern energy systems.
Important Links:
Go Back
