Deep Learning-Driven Knowledge Graph for Power Trading: Analyzing Entity Behavior and Market Factors

The liberalization of China's electricity prices has resulted in a surge of diverse data types in the power trading sector. This study introduces a novel approach leveraging active learning and deep learning to construct a knowledge graph for power trading company evaluation, addressing the challenges of extensive data labeling and enhancing decision-making for power supply and purchase entities. We first use a fine-tuned RoBERTa model as embedding model, and then use deep learning model to extract entities and relationships from unstructured text data to build a knowledge graph. Finally, the constructed knowledge graph is visualized and analyzed in-depth, exploring the relationships between the electricity trading entity, regulatory bodies, and third-party assessment agencies, as well as the impact of market environment, technical standards, and environmental regulations on the decision-making of the trading entity.