Construction and research of forecasting method of electrical fault based on machine learning and decision tree

Background: Machine learning is progressively utilized within the realm of electrical fault detection, enhancing the accuracy of fault discrimination. Objective: This paper aims to utilize five machine learning algorithms to predict faults in three-phase electrical power system and compare the predictive performance of five classical machine learning algorithms. Method: Logistic Regression, Decision Tree, Random Forest, XGBoost, and Support Vector Machine are employed to predict the existence and types of faults in three-phase electrical power system. The algorithms' performance is evaluated by comparing the predictive evaluation metrics. Results: In this paper, Decision Tree exhibited the optimal evaluation metrics, achieving an accuracy of 88% on the test set. Conclusion: The experimental results indicate that Decision Tree exhibits the best performance in predicting faults in power system. This study provides guidance and recommendations for decision-makers in relevant industries.