Transformer Health Index Prediction with Multiple Regression Machine Learning Models
Year: 2026
Type: conference-paper
Abstract: Power transformers are critical assets in electrical power systems, and accurate condition assessment is essential for reliable operation and effective maintenance planning. This paper presents a data-driven machine learning framework for transformer condition assessment based on dissolved gas analysis components and oil quality indicators. Unlike conventional diagnostic methods that rely on predefined thresholds and expert interpretation, the proposed approach integrates data preprocessing, exploratory data analysis, feature selection, model training, and hyperparameter optimization to improve predictive accuracy and generalization capability. A dataset consisting of oil-based diagnostic parameters and the transformer Health Index is analyzed, and ten regression-based machine learning models are evaluated under a unified experimental protocol using both full and selected feature sets. The results indicate that ensemble-based models consistently outperform simpler approaches, demonstrating their effectiveness in capturing complex and nonlinear relationships. Feature selection and hyperparameter tuning are shown to play an important role in mitigating overfitting and improving model robustness on unseen data.
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203714846