Smart Electric Vehicle Charging Management Using Reinforcement Learning on FPGA Platforms

Area of Science:

• Electrical Engineering
• Computer Science
• Artificial Intelligence

Background:

• Growing demand for electric vehicles (EVs) necessitates intelligent charging solutions to manage grid load.
• Existing charging systems often lack the adaptability to handle fluctuating grid conditions and diverse charging priorities.

Purpose of the Study:

• To develop a smart EV charging management system integrating Reinforcement Learning (RL) on a Field-Programmable Gate Array (FPGA) platform.
• To optimize EV charging decisions for grid stress mitigation and prioritized charging needs.
• To create an energy-efficient and scalable solution for modern smart grid environments.

Main Methods:

• Implementation of a Q-learning algorithm for the RL agent on an FPGA.
• Integration of hardware sensors (current, voltage, priority indicators) for environmental perception.
• Dynamic current allocation and task prioritization for multiple EV chargers.
• Hardware design strategies for efficient, real-time operation and low energy consumption.

Main Results:

• The system effectively manages multiple EV chargers, dynamically allocating current and prioritizing charging.
• Demonstrated ability to maintain stable operation under varying demand, improving power efficiency, safety, and service reliability.
• The FPGA-based design ensures real-time response and low energy usage, suitable for embedded applications and potential energy harvesting.

Conclusions:

• The proposed FPGA-based RL system offers a practical framework for efficient EV charging infrastructure.
• The intelligent, sensor-driven approach enables adaptation to grid fluctuations and optimizes energy distribution.
• The scalable design supports expansion for larger installations, contributing to robust smart grid environments.