Partial Discharge Classification and Intelligent Maintenance Decision-Making Based on Spikingformer and LLMs

Partial discharge (PD) classification is an important part of power equipment health management, and accurately identifying PD types and their development trends is crucial for intelligent operation and maintenance of equipment. This article proposes an intelligent PD diagnosis framework that integrates Spikingformer and large language model (LLM). Firstly, event driven PD feature extraction is carried out using Spikingformer combined with spiking neural network (SNN), and the dynamic adjustment mechanism of membrane potential is adopted to improve classification accuracy. Subsequently, based on the SNN classification results, LLM combines device operation data with historical maintenance records to make intelligent operation and maintenance decisions. In addition, we introduce an optimization mechanism based on physics knowledge to enable LLM to consider PD development laws and device physical models when providing decision recommendations, thereby improving the interpretability and adaptability of the system. The experimental results show that this method achieves efficient and accurate PD recognition under various working conditions, and can provide reasonable intelligent operation and maintenance strategies, providing a new solution for equipment health management in fields such as smart grid, transformer monitoring, and wind power operation and maintenance