Using Battery Digital Twins to Elevate EV Performance and Life
Battery digital twins continuously collect battery and vehicle data over the system’s lifetime, accurately estimating the battery’s current and predicting the future state of charge.
Emerging technologies are transforming electric vehicles, improving the driving experience while significantly impacting the supply chain and altering the very strategy OEMs employ to manufacture cars.
Image used courtesy of Adobe Stock
Vehicle systems are being reinvented, including displays, powertrains, infotainment, and telemetry. We are approaching a common industry vision of the software-defined vehicle (SDV) with interconnected systems powerful enough to execute complex software.
Communications using vehicle-to-everything (V2X) technology are possible thanks to integrated software and 5G connectivity, including a crucial link to the cloud. Last but not least, family EVs can offer adaptive safety features for individuals and their driving preferences, while commercial EV fleets can maximize efficiency by deploying smart routing and charging algorithms.
For EV drivers, range anxiety has been a major concern. Coulomb Counting and the Kalman Filter are the standard methods for estimating the battery pack’s state-of-charge (SoC), which provides drivers with a more accurate range estimation—similar to a fuel gauge. As neither method considers the battery's age, each of these SoC estimations becomes increasingly inaccurate throughout the vehicle's lifetime. Moreover, since SoC estimation accuracy forms the basis for all the battery management system’s functional decision-making, a more effective method is necessary—enter the battery digital twin.
Battery Digital Twin
Electra Vehicles, Inc. has pioneered battery digital twin technology development and deployment in collaboration with NXP. Given the high investment and vital role batteries play in EVs, Electra’s Adaptive Battery Digital Twin is deployed as a combined in-vehicle edge-based artificial intelligence (AI) digital model and a remotely computed cloud-based machine learning (ML) battery digital twin. When continuously collecting battery and vehicle parameters over the system’s lifetime, this technology can accurately estimate the battery’s current and predict the future SoC.
To demonstrate how the Adaptive Battery Digital Twin adapts to the battery’s aging over the lifetime of the EV, a comparison was made with the standard Coulomb Counting and Kalman Filter SoC estimation method, see Table 1.
Table 1. Comparing SoC estimation methods—Coulomb Counting and Kalman Filter versus Electra’s Adaptive Battery Digital Twin. Image used courtesy of Electra Vehicles
Aggressive or erratic driving behavior can greatly stress the battery pack, often resulting in battery degradation. As the vehicle ages, the SoC estimation accuracy using the standard methods decreases across various driving styles. The AI/ML model of the battery pack continuously learns from the vehicle’s unique driving patterns, considering battery temperatures due to differing weather conditions, driving loads, and charging regimes to compute the most accurate SoC estimates. This allows the battery management system (BMS) to adjust and control charging conditions and driving limits to extend the battery pack’s range.
But it's not only the battery’s SoC that the digital twin can estimate. Other metrics provide insights into the battery’s state of health (SoH) without physical inspection. Battery anomalies and critical failures can be detected sooner, which can then recommend essential battery maintenance timelines. Providing driving and charging tips to discourage harmful behaviors helps prevent battery damage, which increases the battery pack’s lifespan and enables OEMs to reduce battery pack maintenance and warranty costs. Moreover, having the ability to identify healthy battery packs enables OEMs to offer low-risk battery warranty extension policies.
Deploying Digital Twins
When deployed on a BMS, the Adaptive Battery Digital Twin receives precise voltage, temperature, current, and other relevant data. The data is then synchronized and processed to compute the SoC using an AI algorithm. The resulting SoC measurements are transmitted over CAN to the battery system components, requiring accurate and up-to-date SoC information. The measurements are further processed, compressed, and stored until a secure connection to the cloud can be made via a low-cost internet connection or an over-the-air (OTA) transmission.
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 and then back to the HVBMS for updating. This ensures that the in-vehicle Adaptive Battery Digital Twin is trained with the most up-to-date vehicle and fleet insights and ready to provide continuously improving SoC and SoH estimations.
Figure 1. Electra Vehicles and NXP: Software and Hardware Deployment of the Adaptive Battery Digital Twin. Image used courtesy of NXP
NXP high-voltage BMS (HVBMS) provides the Adaptive Battery Digital Twin with reliable and secure data, processing power, and connectivity to the cloud for efficient EV battery management, see Figure 1. With the S32K3 MCU at the HVBMS’s heart and the S32G GoldBox for computing, onboard storage, and secure cloud connectivity, they work together to produce SoC and SoH measurements of the EV’s battery pack based on the learnings from the cloud-based Adaptive Battery Digital Twin.
Figure 2. Electra Vehicles, Inc. and NXP Semiconductors are accelerating electrification with Edge/AI/ML-based software for EV battery systems. Image used courtesy of NXP
During Embedded World 2023, NXP showed a live demonstration using simulated battery and environmental data on an NXP HVBMS reference design board, which powered the battery digital twin and NXP’s S32K3 chipset. It showed a 12% increase in SoH at year 11 of vehicle life compared to the industry standard vehicle, see Figure 2.
What Does the Future Hold for Software-Defined EVs?
Let’s step back and look at an ideal future for private EV owners, fleet operators, and OEMs.
Imagine an EV that anticipates charging schedules and speeds to extend battery life while allowing for unforeseen road trips and seasonal weather changes. We expect EVs to work with driver assistance and autonomous driving technologies to encourage a more eco-friendly driving style, reducing battery wear and increasing range. Also, to improve the EV resale value, consider having the battery’s SoH and remaining lifetime on the dashboard—as accompaniments to the recorded mileage.
Fleet owners can operate a connected smart fleet that can suggest particular vehicles for daily journeys to distribute battery wear out evenly across the fleet. Imagine receiving daily preventative maintenance schedules based on the specific requirements and predicted remaining lifetime of each vehicle in the fleet.
Finally, OEMs can envision a future where software-defined EVs adjust to each owner’s unique driving styles and charging behaviors dynamically and adjust warranty plans automatically. Based on intelligent AI modeling, imagine new revenue-generating services, like extended warranties, for drivers who are deemed at low risk of having battery warranty claims.
This article was co-authored by Curt Hillier, Technical Director of System Engineering, NXP; Dr. Andreas Both, BMS Software Manager, NXP; Antonio Leone, Director of Business Segment BMS, NXP; Brian Glassman, Ph.D., Director of Product Management, Electra Vehicles, Inc.; and Lanie Meyers, Senior Marketing Manager, Electra Vehicles, Inc.