To implement the people-centered development philosophy, it is necessary to accelerate the construction of an excellent power supply service system, better serve the implementation of national major strategies, and address the urgent and difficult issues in electricity use faced by the public. Following the overall framework of "corpus construction-pre-trained model construction-work order information mining and classification," this study first integrates data from multiple channels, including 95598 service hotlines, WeChat groups, and localized service platforms. Then, by leveraging the capabilities of customer service large language models, it extracts features from work order information to achieve classification of appeal work orders. The results show that the proposed framework enables fine-grained mining and classification of appeal work orders, appeal management featuring mining accurately understanding customers' real needs and emotional states, and providing a foundation for effective appeal handling.