Machine Learning-Based AC Resistance Model for Litz Wire Windings

Optimizing the AC resistance of litz wire windings is crucial for reducing loss and improving the efficiency of energy transfer in power systems. This study presents a novel method for obtaining the AC resistance of litz wire windings by using machine learning algorithms. It simplifies and improves the speed at which AC resistance values are obtained and facilitates the study of AC resistance parameters in litz wire windings. A simulation model is established using an approximate model for optimizing the AC resistance of litz wire windings based on the Dowell equation. Maxwell software is used to generate the training dataset. The frequency, dimensions, number of layers, and conductivity are used as input to the proposed machine learning model, while the AC resistance of litz wire windings is the output variable. A neural network is coupled with the particle swarm optimization method to optimize the network structure to achieve better results. The proposed machine learning model has a prediction accuracy rate of 99%.