Power data entity recognition method based on BiGRU-CRF and NLP

To address the issues of accuracy and efficiency in power data entity recognition, this study proposes a method based on BiGRU-CRF and NLP. A power data dictionary is constructed using hash function initialization, and NLP-based word segmentation is applied. A deep learning model is built by integrating bidirectional gated recurrent unit (BiGRU) with conditional random fields (CRF), enabling efficient and accurate recognition of power data entities. Experimental results show that the design method achieves a maximum processing time of 28.35 ms and an entity recognition accuracy of 96.52% across all entity categories. These results indicate that the designed method not only ensures the efficiency of power data entity recognition but also delivers more accurate recognition performance, meeting the real-time demands of power scheduling and related decision-making.