Dr. Andreas Both, Antonio Leone, Curt Hillier from NXP and Brian Glassman, Lanie Meyers from Electra Vehicles
Drivers, carmakers, and the supply chain share several ambitions for electric vehicles (EVs) – improving range and battery safety while increasing the battery’s lifespan and reducing maintenance costs. Each of these goals can be achieved with new technology enhancements, secure and reliable connectivity, and the innovative use of digital twins using the data the battery produces.
The EV landscape is fast evolving to suit market and performance needs, and advances in battery technology are driving this shift. The cost, performance, and warranty of EV battery packs and sophisticated battery management systems (BMSs) are critical factors influencing EV value. However, it is the BMS’s data with secure cloud connectivity that will help drive these improvements further over the vehicle’s lifetime. But what kind of data does the BMS gather?
Key Functions of the BMS
EV batteries require extremely accurate charging current and output voltages to meet automotive standards. The BMS effectively monitors the battery’s safety, performance, and efficiency, as shown in figure 1. This is achieved by collecting the necessary information to estimate the battery’s state-of-charge (SoC) and state-of-health (SoH). This information is then translated into the EV’s usable driving range – a function equivalent to a fuel gauge. The BMS also ensures that the cells operate within the safe operating area (SOA), which is defined by voltage, current, temperature, and other parameters – operating outside these tight limits can result in battery lifetime degradation, failure, or, worse still, thermal runaway.
Another essential function of the BMS is the management of all the individual cells within the battery pack to ensure a longer lifetime. Cell balancing occurs during charging and while driving, at idle, and in standby. When charging is initiated, the voltage of each cell is measured. During balancing, the highest cells are discharged to the lowest cell voltage to get them to the same SoC. When all the cells are balanced, they can reach their maximum SoC upon completion of charging. The BMS ensures that the charging process is controlled and adapted to the battery’s state.
Estimating the Battery’s SoC
Coulomb counting and the Kalman Filter are the standard methods of estimating the SoC and its accuracy forms the basis for all other BMS functional decision-making. However, these estimations become less accurate over the vehicle’s lifetime because of the accumulated degradation of the battery.
Click image to enlarge
Table 1: Coulomb counting and Kalman Filter estimations of the battery’s SoC become increasingly inaccurate over time
Table 1 shows three driving conditions to demonstrate how the accuracy of these calculations decreases over time: slow steady speed, high steady speed, and quick acceleration and braking. The performance rating for the usual technique drops across these situations as the vehicle ages. Because the SoC is used to estimate range, SoH, and overall lifespan and forms the basis for all other BMS functional decision making, a technology capable of modeling and predicting battery performance is required.
The Adaptive Battery Digital Twin
Electra Vehicles Inc. (Electra) is an artificial intelligence (AI) software innovator focused on maximizing the value of EV battery packs. In collaboration with NXP, the company has developed the adaptive battery digital twin, combining AI and machine learning modeling with physical chemistry battery modeling to provide accurate and predictive outputs. As the vehicle ages, the adaptive battery digital twin learns continuously to provide more accurate SoC and SoH estimations, ensuring more correct BMS controls.
Having looked at what the adaptive battery digital twin can do, let’s consider how Electra Vehicle’s has deployed it on the high-voltage BMS (HVBMS). Powered by NXP’s silicon, the HVBMS provides precise and synchronized high voltage, temperature, and current data. The in-vehicle adaptive battery digital twin, which is deployed on the NXP GreenBox 3 Real-Time Development Platform, utilizes NXP’s S32K3 MCU on the battery management unit (BMU) along with the battery data to estimate the SoC using an AI algorithm that refreshes outputs at a rate of 10Hz. This accurate information is then transmitted over CAN to update the various battery system components, see figure 2.
Click image to enlarge
Figure 2: Electra Vehicles and NXP: adaptive battery digital twin software and hardware deployment
The battery data is filtered and transmitted securely to the NXP S32G GoldBox, where it is further processed, compressed, and stored until a low-cost internet connection becomes available. Alternatively, the data can be compressed for Over-the-Air (OTA) transmission. The NXP GoldBox allows for secure and rapid connection to the Amazon Web Services (AWS) Cloud, which hosts Electra’s adaptive battery digital twin.
After the cloud-based adaptive battery digital twin’s deep machine learning completes its training, a lightweight configuration file in the form of an encrypted OTA update is transmitted back to the EV securely via the GoldBox. From there, the NXP S32K3 MCU is updated. By doing so, Electra’s in-vehicle adaptive battery digital twin is trained with the most up-to-date vehicle and fleet insights and is prepared to give continuously improved SoC and SoH estimations. See table 2 for a comparison of standard and adaptive battery digital twin SoC estimations.
Click image to enlarge
Table 2: Comparing SoC estimation methods shows that the accuracy of the adaptive battery digital twin remains high
During Embedded World 2023 in Nuremberg, Germany, NXP demonstrated simulated battery and environmental data on its high-voltage BMS (HVBMS) reference design board, which power Electra’s adaptive battery digital twin and NXP S32K3 MCU chipset. Two vehicles were shown side by side on a display screen. The first calculated the SoC using the usual Coulomb Counting and Kalman Filter approach, while the second used Electra's adaptive battery digital twin to calculate the SoC and other battery characterization metrics. The simulation showed a 12% increase in battery SoH at year 11 of the vehicle’s life compared to industry standards, see figure 3.
Click image to enlarge
Figure 3: Electra’s adaptive battery digital twin calculates the Dynamic SoC, combined with the highly accurate SoC measurement, leverages the battery’s true performance, resulting in a 12% increase in SoH
The Future of EV Batteries
Future EVs will be more intelligent and able to predict their drivers' requirements. Envisage owning an EV that anticipates charging schedules and speeds to extend battery life while allowing for unforeseen road trips and seasonal weather changes. Expect EVs to work with driver assistance and autonomous driving technologies to reduce battery wear and increase trip range through eco-friendly driving styles.
For fleet owners, consider operating a smart fleet that can suggest certain vehicles for daily journeys to distribute battery wear out evenly across the fleet. Imagine receiving daily preventative maintenance schedules based on the specific requirements of each vehicle and anticipated future usage.
Last but not least, OEMs can see a future when software defined vehicles dynamically adapt to each owner's own driving style and automatically alert the OEM of modification plans. Imagine being able to see insights on possible warranty issues years before they manifest. Consider supplying drivers who are assessed to be at low risk of battery warranty claims with additional revenue-generating services, such as extended warranties. Finally, to boost the resale value of EVs, consider showing the battery's SoH and remaining range on the vehicle's dashboard.
Using Electra’s adaptive battery digital twin approach is crucial to proactively addressing potential accuracy concerns in SoC and SoH estimations. As the number of EVs on the road continues to rise, these will improve consumer perception of the long-term viability of EV brands.