Optimization of Spiral Coil Design for WPT Systems using Machine Learning

This paper presents a spiral coil design for wireless power transfer (WPT) systems using a machine learning (ML)-based optimization method. The designed model allows us to obtain the optimal values of the number of turns N and coil pitch size p, when other coil design parameters such as the coil outer diameter Do, the wire thickness wt, and the operating frequency fs, are given based on the application environment. The proposed ML-based spiral coil design method is assessed by using two metrics: the top-k accuracy and the intersection over union (IoU) factor. The first metric shows that the quality Q factor of the coil, a critical parameter to determine the efficiency of a WPT system, has an error rate less than 4% with respect to the value of the true top-l Q factor in the proposed method. The second metric showcases that the Io U of the proposed method is more than 78% even when a small amount of the available data, less than 1, 000 samples, is used to train the model. The performance of the proposed method is also demonstrated by means of fabricated spiral coils in two application environments. Specifically, for each environment, seven spiral coils are fabricated to find the optimum design point for various values of p and N with fixed values of Do, wt, fs. It is observed that the measured optimal p and N values are identical to the values output by the proposed ML-based optimization method.