Machine learning boost for wireless power transfer
Researchers in Japan have used machine learning to develop a fully numerical technique to improve wireless power transfer (WPT) system design.
WPT systems often struggle with voltage stability when loads change. The researchers at Chiba University used a machine learning-based design method that uses numerical optimization to achieve load-independent operation. This achieved a stable output voltage with fluctuations under 5% and a high efficiency of 86.7% under varying loads.
Recently, load-independent (LI) operation has been attracting attention to keep the output voltage stable and maintain zero-voltage switching (ZVS) even when the load changes. However, achieving LI operation requires very precise circuit component values for the inductors and capacitors, which are typically calculated using complex analytical equations. These equations are often based on idealized assumptions and do not capture real-world complexities.
The approach developed by Professor Hiroo Sekiya from the Graduate School of Informatics at Chiba University used differential equations that capture how voltages and currents evolve over time within the system, taking into account real-world component characteristics.
These equations are then solved numerically, step by step, until the system reaches steady-state conditions. An evaluation function assesses the system’s performance based on key objectives such as output voltage stability, power-delivery efficiency, and total harmonic distortion. A genetic algorithm then updates the system parameters to improve the evaluation score, and the process is repeated until the desired load-independent operation is achieved.
“We established a novel design procedure for a LI-WPT system that achieves a constant output voltage without control against load variations. We believe that load independence is a key technology for the social implementation of WPT systems. Additionally, this is the first success of a fully numerical design based on machine learning in the field of power electronics research,” said Prof. Sekiya.
The team used the method with a class-EF WPT system, which combines a class-EF inverter with a class-D rectifier. The traditional class-EF inverter without LI operation can maintain ZVS only at its rated operating point. If the load changes, ZVS is lost, and efficiency drops. In contrast, the LI version designed by the team kept ZVS and output voltage stable even when the load varied.
In the conventional system of the LI inverter, the output voltage could fluctuate by as much as 18% when the load changed. In contrast, the system designed with the fully numerical method kept this variation below 5%, demonstrating significantly greater stability. By accurately accounting for the effects of diode parasitic capacitance, the new approach also maintained better performance at light loads. A detailed power-loss analysis revealed that the transmission coil dissipated nearly the same amount of power across different load conditions, thanks to the system’s ability to keep the output current steady.
At its rated operating point, the LI class-EF WPT system achieved a power-delivery efficiency of 86.7% at 6.78 MHz and delivered more than 23 W of output power.
“Due to LI operation, the WPT system can be constructed in a simple manner, thereby reducing the cost and size. Our goal is to make WPT commonplace within the next 5 to 10 years,” says Prof. Sekiya.
ML-Based Fully-Numerical Design Method for Load-Independent Class-EF WPT Systems: 10.1109/TCSI.2025.3579127
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
