Machine learning methods for load losses calculation in power transformers

Load losses in power transformers are a critical factor influencing both energy efficiency and long-term operational costs. Accurate estimation of these losses is essential during the design phase, as it directly affects transformer performance, sizing, and compliance with energy efficiency standards. Traditionally, load loss calculations rely on analytical models and experimental testing, which can be time-consuming, resource-intensive, and sometimes limited in their ability to generalize across different transformer configurations. This thesis explores the application of machine learning techniques to improve the prediction of load losses using only transformer design and performance data. The objective is to develop a flexible and data-driven approach that can complement or partially replace conventional methods, offering faster and potentially more accurate estimations. By training supervised learning models on datasets containing detailed design features and measured performance metrics, the study aims to capture the underlying patterns and interactions that influence load losses. The proposed methodology offers a promising direction for enhancing the efficiency and scalability of transformer design processes, and opens opportunities for further integration of machine learning in the electrical engineering domain.