Machine Learning-Aided Optimization Strategy of DGS Resonators for WPT Systems
Year: 2026
Type: article
Abstract: The features of defected ground structures (DGSs) make them well-suited for compact resonator design, which is an essential part of near-field wireless power transfer systems. Based on its geometry, the defect etched on the ground plane defines the circuit behavior, including the resonant frequency. Therefore, this work presents a machine learning (ML) approach for modeling DGS-based resonators and predicting their resonant behavior. The square defect, whose size and location were adjusted to explore all possible outcomes, was used in the designed DGS resonator with a 25 × 25 mm2area. Implementing defect modifications facilitates the collection of approximately 20,000 datasets used for training, validating, and testing the ML model. In particular, the random forest (RF) technique is used to build a predictive model because it can quickly solve regression-type problems with high accuracy. Essential hyperparameters are carefully selected to enhance the model’s accuracy during training, and its performance is evaluated through key performance metrics, i.e., mean absolute percentage error (MAPE), explained variance score, mean absolute error, and R-squared. MAPE of 0.32% validates the effectiveness of the RF-based model in predicting resonant frequencies. Furthermore, RF’s performance is compared with similar ML models to benchmark the attained results. The resonators anticipated by the RF model are validated through measurements.
Institutions: Nazarbayev University, University of Sharjah
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W7213361356