Green Digital Operation and Maintenance Technology of Power Equipment Based on Deep Learning

Abstract

In order to solve the problem of low accuracy in traditional digital operation and maintenance of power equipment, the author proposes a research on green digital operation and maintenance technology of power equipment based on deep learning. The author first analyzed the text characteristics and segmentation difficulties of work orders, summarized seven types of entities, and manually annotated 3452 work orders to form a training set, Secondly, pre train the BERT module using relevant equipment testing and fault analysis reports to obtain power word vectors, Then use the BiLSTM module to predict entity labels, Finally, the CRF module was introduced to optimize the prediction labels and conduct Chinese entity recognition experiments on 1000 work orders. The experimental results indicate that: Compared with LSTM, BiLSTM, and BiLSTM-CRF models, the BERT BiLSTM-CRF model has better entity recognition performance in power primary equipment operation and maintenance work order texts, with an F1 value of 85.7\%, which is 11.6%, 7.3%, and 4.8% higher than the other three models, respectively. This system can effectively improve the work efficiency of operation and maintenance personnel, reduce their communication costs, achieve efficient management of on-site equipment, and assist enterprises in achieving production goals of improving quality and efficiency.