Knowledge-driven recognition methodology for electricity safety hazard scenarios

Abstract

This study proposes a knowledge-driven recognition methodology for electrical safety hazard scenarios to address the inability of maintenance personnel to accurately identify safety hazards and deal with them effectively due to complicated electrical safety hazard scenarios. First, a BERT-BiLSTM-CRF-based safety hazard entity recognition model and ResPCNN-ATT-based safety hazard relation extraction model are designed. The entity recognition and relation extraction in electricity safety hazard records are implemented by them. Then, the knowledge graph of the electricity safety hazard domain is constructed, and the semantic information of scenes provided in the hazard records is incorporated into the relationship constraints among various entities. Finally, an intelligent reasoning model of hazard scenarios based on a knowledge graph of electricity safety hazards is proposed. The hazard scenes and solution measures can be quickly located based on the types of hazard equipment and hazardous phenomena. The proposed approach’s effectiveness and engineering practicality was verified using the accumulated hazard records of a provincial power company.

Keywords

• Electricity safety hazards
• Domain knowledge graph
• Deep learning
• Knowledge driven
• Hazard scenes identification