کنفرانس بین المللی ماشین و محرکه های الکتریکی

صفحه اصلی / The 5th International Conference on Electrical Machines and Drives

Neural Network-Based Prediction of AC Losses and Performance in High-Speed PMSM with Hairpin Winding

نویسندگان :

Mohammad Sepehr Zamani¹ Aghil Ghaheri² Ebrahim Afjei³

1- Faculty of Electrical Engineering, Shahid Beheshti University, Tehran, Iran 2- Faculty of Electrical Engineering, Shahid Beheshti University, Tehran, Iran 3- Faculty of Electrical Engineering, Shahid Beheshti University, Tehran, Iran

کلمات کلیدی :

High-speed PMSM،hairpin winding،neural network،AC loss،hybrid modeling،prediction،machine learning

چکیده :

This study presents a hybrid modeling approach that combines 2-D Finite Element Method (FEM) simulations with a Neural Network (NN) to predict key parameters of a high-speed Permanent Magnet Synchronous Machine (PMSM), including DC losses, AC losses of individual hairpin winding layers, skin effect losses, and efficiency. These parameters are critical for designers to understand the impact of each layer’s height on performance and losses. The methodology consists of two phases: generating training data from 2-D FEM simulations and developing a NN whose configuration and parameters are optimized via Bayesian optimization to ensure high accuracy. The developed NN demonstrates excellent predictive performance on the test dataset, achieving R² values above 99% for efficiency, DC losses, and skin-effect losses, and over 99.4% for AC losses in each hairpin winding layer. Accuracy is particularly high for layers near the air gap, which have the greatest impact on overall performance. The results confirm that the NN effectively captures the relationship between hairpin winding layer heights and output parameters in high-speed PMSM with hairpin winding, which enables reliable prediction of performance and AC losses.