Research on Matching Texts of Grid Equipment Defects with Technical Supervision Regulations Based on BERT

Frequent defects in grid equipment pose a significant threat to the operational safety of the grid and the stability of the power supply. Therefore, it is crucial to implement corrective measures swiftly and accurately in accordance with the relevant technical supervision regulations. However, the traditional approach of relying on information and empirical judgment is often inefficient and prone to errors. To intelligently match defective text with technical supervision regulations more effectively, this paper designs a semantically similar matching task between defective grid text and technical supervision regulations based on the BERT pre-training model. It also compares and analyzes the performance of BERT and its improved model in this task by collecting and expanding defective text data from 14 types of grid equipment. The experimental results indicate that the average accuracy of the ERNIE model in matching can reach 89.84%. This model effectively captures the semantic relationships between defective text and technical details, providing intelligent support for technicians and enhancing the accuracy and efficiency of operation and maintenance management for grid equipment defects.