Wireless power transfer (WPT) technology is rapidly evolving, promising a future where charging cables and connectors become obsolete. But while WPT systems are already used in everything from smartphones to electric toothbrushes and industrial sensors in IoT, they often struggle with voltage instability under varying load conditions—a critical barrier to widespread adoption.
Now, researchers from Chiba University and partner institutions have introduced a machine learning-based design approach that could dramatically enhance the efficiency and reliability of wireless charging systems. Their work enables load-independent wireless power transfer with voltage fluctuations of less than 5% and efficiency rates up to 86.7%, marking a significant advancement in power electronics and circuit design.
What is WPT, and why is stability a challenge?
Wireless power transfer systems deliver electricity through electromagnetic fields rather than physical wires. A standard WPT setup includes a transmitter coil connected to a power source and a receiver coil attached to an electronic device. When properly aligned, the transmitter generates a magnetic field that induces a current in the receiver, wirelessly powering the load.
However, maintaining a stable output voltage becomes challenging as the power demand (load) changes. Traditional WPT designs often lose efficiency and zero-voltage switching (ZVS)—a condition critical for reducing power loss and electromagnetic interference—when operating outside of tightly controlled parameters.
Introducing load-independent WPT using ML
To address this, a team led by Professor Hiroo Sekiya at Chiba University developed a machine learning-based numerical optimization method for WPT circuit design. The goal? Achieve load-independent (LI) operation, where voltage remains stable across a wide range of load conditions without the need for real-time control systems.
Key Features of the Machine Learning-Based WPT Design:
- Fully numerical design using real-world component models (including parasitic effects).
- Genetic algorithm optimization to find the best circuit parameters for stability and efficiency.
- Evaluation functions that consider voltage regulation, power delivery efficiency, and harmonic distortion.
- Differential equation modeling to simulate realistic behavior over time.
This method is the first of its kind to use AI-based optimization for designing a practical WPT system, pushing the boundaries of automated circuit design in power electronics.
Real-world testing: class-EF WPT System
The research team applied their design to a class-EF WPT system, which combines a class-EF inverter with a class-D rectifier—a setup often used in high-frequency applications.
Traditional vs. ML-Optimized WPT:
- Traditional design: ZVS is maintained only at the rated load; voltage fluctuates up to 18% with load changes.
- Machine learning-based design: Stable voltage with fluctuations under 5%; ZVS maintained across all load levels.
The optimized system also handled light-load conditions better by compensating for diode parasitic capacitance, a commonly overlooked factor in conventional designs.
At 6.78 MHz, the system achieved:
- 86.7% power-delivery efficiency
- Over 23 W of output power
- Minimal power loss across varying loads, thanks to consistent output current
Why this matters: toward practical, cost-effective WPS
According to Prof. Sekiya, this breakthrough could help pave the way toward a fully wireless society:
“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.”
Beyond technical performance, simplified WPT circuit design means lower costs, smaller form factors, and easier integration into consumer and industrial applications.
The future of WPT and AI-Driven Circuit Design
The success of this machine learning-driven method not only sets a new standard for efficient wireless energy transfer, but also showcases the broader potential of AI in electrical engineering.
With applications ranging from wireless charging stations for electric vehicles (EVs) to medical implants and smart sensors, this innovation could redefine how engineers design power systems in the next decade.
The researchers aim to continue refining their technique and advocate for widespread deployment of WPT technology within 5 to 10 years—a future where charging cables are a thing of the past, and AI-designed circuits power our everyday lives.
Read next: New guide to improve building safety across the UK – Build in Digital
Are you a building professional? Sign up for a FREE MEMBERSHIP to upload news stories, post job vacancies, and connect with colleagues on our secure social feed.
