Real-Time Temperature Distribution Estimation in Permanent Magnet Synchronous Motors (PMSMs) Using Machine Learning: A Path Toward Energy Efficiency and Sustainability

Abstract This research investigates real-time temperature distribution estimation in Permanent Magnet Synchronous Motors (PMSMs), leveraging advanced supervised machine learning techniques to overcome the limitations of conventional measurement methods. Accurate temperature estimation is critical for optimizing motor performance, enhancing energy efficiency, and ensuring system reliability, particularly in applications geared toward sustainable energy systems and decarbonization. A range of regression algorithms—including Linear Regression, Random Forest Regressor, XGBoost Regressor, Gradient Boosting Regressor, and more—were explored to develop predictive models with high accuracy. Among these, the XGBoost Regressor emerged as the best-performing model, achieving an R² score of 0.9810, a Mean Absolute Error (MAE) of 1.7767, and a Root Mean Square Error (RMSE) of 2.6227. Its effectiveness is further validated by its bell-shaped error distribution plot, demonstrating strong generalization and minimal overfitting. To enhance accessibility and scalability, a web application was developed to deploy the trained model. This web-based solution provides real-time temperature predictions based on motor input parameters, enabling professionals to monitor and manage PMSM functionality effectively. The application is particularly useful in energy-critical operations, offering a practical tool for reducing energy losses, prolonging motor lifespan, and advancing system efficiency. The integration of machine learning with PMSM diagnostics highlights the potential of artificial intelligence in energy-efficient motor control and decarbonization. Beyond the technical accomplishments, this research lays a foundation for future advancements in machine learning-driven motor diagnostics, energy optimization, and sustainability practices.