Hot spot temperature modeling in power transformers using variation in load and ambient temperature

Abstract

Hot spot temperature is the most significant factor influencing the aging and reliability of oil-immersed power transformers, as it accelerates insulation deterioration and reduces operational longevity. Maintaining reliability of the system, minimizing thermal failures, and ensuring safe transformer loading depend on precisely estimating this temperature. This paper introduces a real-time applicable, data-driven approach that estimates transformer hot spot temperature using only historical low-voltage (LV) load and ambient temperature data,readily available without requiring additional effort or special measurements. More flexible and adaptable than existing models, the suggested approach is based just on actual operational data and does not depend on limiting assumptions. This model provides a practical and assumption-free solution for utility-level transformer monitoring and decision-making by using machine learning approaches to record the thermal behavior of the transformer under different scenarios.