Machine Learning–Based Short-Circuit Prediction in Electricity Distribution Transformers

Electricity transformers are critical components of electrical power systems, ensuring stable voltage regulation and reliable power distribution. However, unexpected failures, particularly due to short circuits, can lead to significant power outages and costly repairs. Traditional methods used for transformer maintenance rely on scheduled inspections and historical fault data, which often fail to detect faults in time to prevent major failures. This study explores the use of machine learning to predict short-circuit failures in transformers, enabling a more efficient and proactive maintenance regime. Data from two transformers, one in good condition and another that exhibited signs of failure, were used for model training. Key parameters such as load, oil temperature, dissolved gas levels and current were analysed to develop predictive maintenance models. Several machine learning models, including Random Forest, Linear Regression, Support Vector Machine and Decision Tree Regression, were compared based on their predictive accuracy using metrics such as mean-squared error (MSE) and R-squared (R²). The results show that Random Forest model has the highest accuracy of 99.8%. The implementation of a user-friendly dashboard further enhanced data visualization and could potentially facilitate actionable insights for operators. This research underscores the significant potential of machine learning to enhance transformer reliability, prevent unexpected failures and reduce maintenance costs, ultimately contributing to a more resilient and efficient power infrastructure.