Vehicle Position Detection Based on Machine Learning Algorithms in Dynamic Wireless Charging.

Area of Science:

• Electrical Engineering
• Computer Science
• Automotive Technology

Background:

• Dynamic wireless charging (DWC) addresses electric vehicle range anxiety by enabling charging during motion.
• Segmented coil arrays are preferred for DWC due to higher efficiency and lower electromagnetic interference.
• Accurate vehicle positioning is crucial for DWC systems to activate/deactivate transmitter coils efficiently.

Purpose of the Study:

• To evaluate machine learning algorithms for precise electric vehicle position detection in DWC systems.
• To assess algorithm performance across varying vehicle ground clearances and speeds.
• To identify the most accurate machine learning model for real-time DWC coil management.

Main Methods:

• Tested eight distinct machine learning algorithms for vehicle position determination.
• Evaluated algorithm accuracy based on predicted versus actual vehicle positions.
• Compared performance metrics including prediction error for different ground clearances and speeds.

Main Results:

• The random forest algorithm demonstrated superior performance in predicting vehicle positions.
• This algorithm achieved the lowest error rate among all tested models.
• Performance was consistent across various vehicle speeds and ground clearances.

Conclusions:

• Machine learning, particularly the random forest algorithm, offers a robust solution for precise vehicle localization in DWC systems.
• Accurate positioning enables efficient energy transfer and management of transmitter coils.
• This approach enhances the viability and reliability of dynamic wireless charging for electric vehicles.