Research on Gradual Fault Prediction of Current Transformers Based on Machine Learning Algorithms

Current transformers (CTs) are critical devices in power systems. Their gradual faults (such as insulation aging and core loss) are often hidden and progressive, making them difficult to detect and predict using traditional methods. This paper proposes a gradual fault prediction model for CTs based on machine learning algorithms. By constructing high-dimensional feature vectors, the model extracts, analyzes, and models CT fault characteristics and uses optimized machine learning algorithms to effectively predict gradual faults. First, the study extracts typical fault features from historical operational data of CTs using feature engineering methods and enhances the representation capability of these features through time-series analysis. Subsequently, various machine learning algorithms, including support vector machines (SVM), random forests (RF), and neural networks, are employed to train the fault prediction model. The model’s prediction accuracy and robustness are further improved through ensemble learning strategies. Experimental results show that the proposed model significantly improves the prediction accuracy of gradual faults in CTs, providing better robustness and generalization compared to traditional methods. This research can serve as a technical support for the safe and stable operation of power systems. Finally, the paper discusses the model’s application scenarios and deployment strategies in real-world power systems, providing insights and references for future studies.