Research on Quality Assessment and Report Generation of Power Grid Fault Datasets Based on Prompt Engineering

This study proposes an automated quality assessment method for power grid fault datasets based on large language models (LLMs) and prompt engineering, along with the automatic generation of quality assessment reports. By leveraging the semantic understanding capabilities of LLMs and the flexibility of prompt engineering, the method efficiently analyzes and accurately detects multidimensional issues such as missing values, duplicate entries, outliers, and irregular time intervals within the datasets. Experimental results demonstrate that the prompt-engineering-based approach achieves an overall accuracy of 86% in data quality assessment and 100% accuracy in outlier detection. Furthermore, the method can automatically generate detailed assessment reports in Word format, including visual analytical charts. This approach enhances the automation and accuracy of dataset quality assessments, providing reliable data support for intelligent operation and maintenance of power systems.