The hazards caused by electrical fires are receiving increasing attention, and the largest proportion of their causes is arcing faults. Arcing is usually caused by damage or overload of electrical components, which in turn may lead to damage to electrical equipment and start a fire. Arc identification is an important preventive technology to monitor arcing in electrical equipment so that timely countermeasures can be taken, and is an important part of smart electricity. In this paper, we conduct a study on the arc fault identification method, firstly, we build an experimental platform according to the national standard, then analyze the arc characteristics of different household appliance load combinations and perform feature extraction; then we propose an arc identification method based on CatBoost classification model, and use the CatBoost model to train the extracted features to achieve fast identification of arc accidents; after test set It is verified that the proposed arc recognition method based on CatBoost classification model has higher accuracy and recall rate compared with existing recognition classification methods such as SVM and Random Forest, which can effectively improve the recognition accuracy of arcing accidents.