A Method for Extracting Risk Information of Scheduled Maintenance in Power Grids Based on Universal Information Extraction

To enhance fault prevention and handling guidance capabilities in the context of scheduled maintenance for power grids, this paper proposes a method for identifying entities in power dispatch texts based on a Universal Information Extraction (UIE) model. Focusing on risk forms for scheduled maintenance in power grids, this paper proposes an entity annotation method for extracting key information from these forms. By leveraging the general entity extraction capabilities of the Universal Information Extraction (UIE) model, we identify and output structured data from the unstructured text in the risk forms. This unified representation of textual entities enhances the representation capabilities of entity information. Utilizing the ERNIE 3.0 network to learn the contextual semantic features of sequences, we identify key information such as power grid equipment names, fault and anomaly types, and grid risk conditions. Experimental results indicate that the general information extraction model based on UIE achieves an F1 score of 95.78% on the scheduled maintenance risk form dataset. Validation with actual data from a specific power grid demonstrates that the proposed method has higher accuracy in power dispatch entity recognition compared to other algorithms, providing robust online support for power dispatch business scenarios.