EV battery charging technology has always had to find the right balance between charging speed and battery longevity. If the charging speed is too fast, it wears down the battery. If the charging is too slow, nobody is happy.
Researchers Meng Yuan from Victoria University of Wellington and Changfu Zou from Chalmers University of Technology in Sweden may have cracked this long-standing problem using an AI technique called deep reinforcement learning, and the results are pretty encouraging.
Their study, published in IEEE Transactions on Transportation Electrification, introduces a new AI charging system that learns to charge a battery quickly while actively protecting its long-term health.
What’s different about this new AI-based charging system?
The system uses a machine learning method called TD3, which is a fancy way of saying the AI learns by trial and error across thousands of simulated charging sessions. What makes it different is that it adapts its charging strategy based on how degraded the battery already is.
Most traditional chargers use a fixed routine. They start with full power and taper off as the battery reaches its capacity. The problem is that this routine doesn’t care whether the battery is brand new or has been through hundreds of charge cycles.
The new AI-based charging strategy sidesteps this by learning a relationship between the battery’s health and the maximum safe charging voltage, then using that information to make smarter decisions in real time.
Does the battery last longer?
In simulations using a real-world battery model, the proposed method extended battery life by nearly 23% over standard charging methods, reaching 703 equivalent full cycles compared to just 572 with conventional charging. Charging time also remained competitive at around 24 minutes for an 80% charge.
The team trained the entire system on a consumer-grade desktop with an Intel i5 processor and an NVIDIA RTX 3060 GPU. As the researchers note, this “demonstrates that the proposed framework can be effectively trained on widely available hardware without needing access to specialized high-performance computing clusters.”
It is still early days, and the method needs testing outside of simulations. But if it holds up, smarter charging could quietly become one of the biggest upgrades your next EV gets.