Electric vehicle (EV) charging is getting a jolt with an innovative new AI algorithm that boosts efficiency, reduces cost, and keeps the grid from short-circuiting under pressure. Developed by a team of researchers from the Royal Military College of Canada (RMC), the real-time smart solution optimizes charging schedules for large parking lots, balancing quick charging with energy availability. By making charging faster, cheaper, and more available, the AI-powered algorithm could help pave the way for more widespread adoption of EVs—a cleaner option for reducing emissions and meeting climate goals.
“Optimizing the charging schedule of EVs in a smart parking lot has huge impacts not only on the consumers, who end up paying a lower price, but also on the environment when electricity usage is maximized during periods when it is plentiful,” said Vincent Roberge, study lead author and a professor in the Department of Electrical and Computer Engineering at RMC.
While the popularity of EVs is growing, one major downfall is the availability of vehicle charging stations. Anticipating and managing the technology’s reliance on the power grid is key to keeping drivers happy and energy infrastructure stable. This is especially important in large parking lots, where hundreds of drivers need to charge their vehicles simultaneously.
To address this, the researchers developed an AI-powered algorithm that optimizes charging schedules based on vehicle arrival and departure times, charging time, energy demand, electricity cost based on time of day, and charging rate limits. Using this data, and accounting for all cars in the parking lot, it calculates different combinations of charging schedules, picking the best option based on minimizing cost while avoiding overloading the power grid.
The researchers simulated different EV parking lot sizes to test the algorithm’s performance. They started with a small 20-EV parking lot and then scaled up the model to parking lots with 40 to 500 vehicles.
The team developed the algorithm using two NVIDIA RTX A6000 GPUs awarded through the NVIDIA Academic Grant Program. It uses a particle swarm optimization (PSO) algorithm, boosted by NVIDIA CUDA-accelerated GPU parallel processing for automated, real-time updates as vehicles enter or exit the lot. According to Roberge, the researchers used PSO, an AI technique from the field of computational intelligence, to compute optimized schedules for EV recharging.
“The PSO works by independently improving a large number of possible solutions. These can be evaluated in parallel on the GPU, greatly reducing the time needed to perform the optimization,” Roberge said.
The model runs on multicore CPUs and GPUs, achieving real-time performance with an NVIDIA GeForce RTX 4070 Ti GPU. CUDA-accelerated GPUs greatly boost the scheduling process, delivering a speedup of up to 247.6x, optimizing charging for a 500-EV parking lot in under 30 seconds.
By scheduling EV charging during off-peak hours, the model can help reduce strain on the electrical grid and cut reliance on fossil-fuel power plants, lowering emissions. Optimized charging schedules could also ease the need for costly infrastructure upgrades, improve grid stability, and maximize charging capacity by reducing peak power demand and avoiding periods of high-cost energy use.
The researchers are exploring additional applications of CUDA and GPUs for large-scale smart grid optimization. They’re working on reconfiguring the power distribution network to accommodate renewable energy resources.
“This reconfiguration will ensure that the distribution network is always operating in an optimized state no matter the demand in energy or variation in the renewable energy production,” Roberge said.
Read the research Parallel Algorithm on Multicore Processor and Graphics Processing Unit for the Optimization of Electric Vehicle Recharge Scheduling.