Electric and hybrid cars passing almost silently by are a realized science fiction. Though they are now almost quotidian, they are far from ubiquitous. Efficient and sustainable transportation is becoming increasingly imperative as emissions drive climate change and fossil fuels are depleting. Driving new innovations to improve the efficiency and practicality of electric vehicles is the cornerstone to their wider adoption by the everyday commuter.
The research of Darshika G. Perera, UCCS Professor, and Anne K. Madsen, UCCS Ph.D. graduate, has achieved just such an innovation.
At the heart of electric vehicles is the battery that drives them. Perera and Madsen created a new hardware accelerator to govern batteries, particularly lithium-ion batteries. This new technology will increase both the efficiency and output of batteries and is ready for implementation into the automotive industry. “Nobody is doing anything like this so far,” said Perera.
There are currently two models for controlling the charge and discharge of these batteries, a Model Predictive Controller (MPC): an equivalent circuit model and a physics-based model, but up till now the physics-based model has remained theoretical. Unlike the equivalent circuit model, the physics-based model looks at the chemical reaction inside the battery. Utilizing the physics-based model of the battery allows the control system to operate on the chemical and physical process of the battery that are the root cause of battery degradation.
Previous research conducted by Trimboli, Plett and their students presented a theoretical approach to implementing computationally intense algorithms in resource constrained systems for a physics-based model (PBM). However, the high computational complexity of PBM based MPC, and the necessary physics-based observer, in this case an extended Kalman Filter, has prevented these techniques from being realized on resource-constrained embedded devices, making them infeasible for portable battery packs and battery management systems.
Because the physics-based model predictive controller (PB-MPC) is far more computationally intensive than the equivalent circuit model that is currently used, it has not been utilized in a real world scenario despite its surpassing efficiency.
In their research work, Perera and Madsen introduce a novel, unique and efficient field programmable gate array (FPGA) based embedded hardware accelerator for PB-MPC smart sensor for battery cell management, specifically on embedded devices, by addressing the computational complexity of physics-based control.
“We discussed the idea and she was excited about it because electric vehicles will be the future. We are trying to bring electric vehicles to green energy,” said Perera.
The physics-based model directly determines the chemical reactions of the battery, improving its useful life and increasing the effectiveness of its charge. This hardware accelerator developed by Perera and Madsen increases the speedup of a battery and reduces the time it takes to run a software task and the task execution in the hardware.
It is currently the only model of its kind.
Perera and Madsen developed a hardware and software system to predict charging and discharging, when to charge and when to discharge, and how much to charge and how much to discharge. “We came up with a brand new hardware architecture,” said Perera.
Utilizing hardware based on FPGAs to support and accelerate computationally complex PB-MPC, Perera and Madsen achieved substantial speedup compared to its software counterparts. With these speedups, the embedded PB-MPC hardware smart sensor is able to monitor and control multiple battery cells individually. These speedup results also indicate that FPGAs could accommodate more computationally intense versions and configurations of the physics-based model, and still perform within the required timeframe and safety margins.
Utilization of a physics-based model enables closer observation as well as tracking and controlling of the chemical aging mechanisms in lithium-ion batteries, which is otherwise disregarded when utilizing equivalent-circuit model.
With a FPGAs, which are faster, a large number battery cells and battery packs can be managed. Potentially, a single FPGA could manage every cell, rather than a microprocessor on each battery cell or battery pack. This would be both space and cost efficient.
This work expands the realm of the possible for using more accurate but computationally intense methods on resource constrained systems. The unique capabilities of FPGAs can open the door to more effective and efficient smart systems.
The proposed embedded PB-MPC hardware accelerator achieved 58 times the speedup compared to its embedded software counterpart, while maintaining the small footprint required for portable systems. This speedup enables more battery cell management utilizing a single chip compared to that of embedded processor-based solutions.
Not only does this single chip have a small footprint, but it requires less power and is more speed efficient. Using a single programmable chip is also versatile. It would allow for new programming without new hardware and can be utilized for other tasks when it is not being used for battery management. Using a single chip instead of the many required for the current model has future potential for growth and efficiency.
This new FPGA-based battery management system will increase the efficiency and power of the batteries driving electric and hybrid vehicles. As energy-efficient modes of transportation become ever more imperative for the future of the automotive industry, the individual consumer and the environment, improving the viability of electric and hybrid vehicles is crucial work. The impact on the automotive industry would be an impact for the world.
While the research was done in digital simulations and FPGA prototyping, the PB-MPC is ready for real world application if funded by the automotive industry.
Perera is currently researching efficient computational modes for edge-computing platforms, utilizing machine learning to enhance the cognitive ability of cognitive radio as well as neuromorphic, brain-inspired, computing. “I need to challenge myself and challenge my students too,” said Perera. After receiving her Ph.D. from UCCS, Madsen is now a researcher with Space Force.