Large Language Model-Based Operation Ticket Generation: A Case Study

The power dispatch task is becoming increasingly complex since more and more factors need to be considered, such as weather conditions, energy storage, and the electricity market. To ensure the safe operation of power dispatch, the operation ticket is utilized as a highly specialized record of operation commands to instruct the execution of dispatch operations. Currently, the operation ticket generation mainly adopts the rule-based method, which heavily relies on human expertise and is both time-consuming and cost-expensive. To address these challenges, this paper proposes a Large Language Model (LLM) based operation ticket generation framework, which leverages the natural language understanding and instruction-following capabilities of LLM to generate operation tickets. The framework explores both In-Context Learning and Supervised Fine-Tuning methods and is validated using real dispatch ticket data from the China Southern Power Grid Company. Empirical results demonstrate that the proposed framework achieves an accuracy of 98%.