New Model Boosts Accuracy of EV Battery Charge Estimation
In the rapidly evolving world of electric mobility, one of the most critical challenges remains the precise estimation of a lithium-ion battery’s state of charge (SOC). Accurate SOC monitoring is not merely a technical detail—it directly impacts vehicle range, safety, battery longevity, and overall user confidence in electric vehicles (EVs). Despite significant advancements in battery technology and management systems, achieving high-precision SOC estimation under dynamic driving conditions continues to challenge engineers and researchers globally.
Now, a breakthrough study conducted by Wu Shengli and Guo Qi from the School of Traffic and Transportation at Chongqing Jiaotong University, in collaboration with Xing Wenting from the School of Management Science and Engineering at Chongqing Technology and Business University, has introduced a novel approach that significantly improves both the accuracy and computational efficiency of SOC estimation. Published in the September 2024 issue of Energy Storage Science and Technology, their research presents a fractional variable resistance-capacitance (FVOM) model combined with an enhanced filtering algorithm, setting a new benchmark for real-time battery state prediction.
The significance of this work lies in its ability to address long-standing limitations in traditional battery modeling techniques. Conventional equivalent circuit models (ECMs), widely used for SOC estimation, rely on combinations of resistors and capacitors to simulate the dynamic behavior of batteries. While simple models like the Rint or Thevenin configurations are computationally efficient, they often lack the precision needed under complex load conditions. On the other hand, higher-order models with multiple RC networks can achieve better accuracy but at the cost of increased complexity and computational burden—making them less suitable for onboard implementation in real-world EVs.
This trade-off between model fidelity and practicality has been a persistent bottleneck in battery management system (BMS) design. As electric vehicles face increasingly variable driving patterns—from stop-and-go city traffic to rapid acceleration on highways—battery models must adapt quickly and accurately to transient changes in current and temperature. The inability to do so can lead to inaccurate SOC readings, which may result in unexpected shutdowns, reduced driving range, or even overcharging and thermal runaway risks.
Recognizing these challenges, the research team turned to fractional calculus, a mathematical framework that extends traditional integer-order derivatives and integrals to non-integer orders. Unlike classical models that assume idealized exponential responses, fractional-order models can capture the distributed and memory-dependent characteristics of electrochemical processes within lithium-ion batteries more naturally. This includes phenomena such as ion diffusion in porous electrodes and surface charge transfer, which exhibit power-law dynamics rather than simple exponential decay.
However, while prior studies have explored fractional-order models, many adopt fixed-order structures, assuming that the battery’s dynamic behavior remains consistent across different states of charge. This assumption, the authors argue, does not reflect reality. Lithium-ion batteries exhibit strong nonlinearity, especially at low and high SOC levels, where electrochemical polarization and concentration gradients change dramatically. A one-size-fits-all model, therefore, cannot optimally represent the full spectrum of battery behavior.
To overcome this limitation, Wu, Guo, and Xing proposed a variable-order approach—what they call the Fractional Variable Resistance-Capacitance (FVOM) model. Instead of using a single fixed fractional order, their model dynamically adjusts its structure based on the current SOC. Specifically, the model switches between a first-order fractional RC configuration (FOM-1RC) and a second-order version (FOM-2RC), depending on which provides the best fit at a given SOC level.
The key innovation lies in how the optimal model order is selected. Rather than relying on heuristic rules or manual tuning, the team employed the Akaike Information Criterion (AIC), a statistical tool designed to balance model accuracy against complexity. By calculating the AIC value for both FOM-1RC and FOM-2RC models across different SOC ranges, they identified the configuration that delivers the highest predictive accuracy without unnecessary computational overhead.
Their findings revealed a clear pattern: in the mid-SOC range (between 10% and 90%), the simpler FOM-1RC model performs nearly as well as the more complex FOM-2RC, making it the preferred choice for balancing efficiency and precision. However, at the extremes—below 10% and above 90% SOC—the battery’s behavior becomes highly nonlinear, and the additional dynamics captured by the FOM-2RC model are essential for maintaining accuracy.
This adaptive switching mechanism allows the FVOM model to remain lightweight during most of the battery’s operating range while still delivering high fidelity when it matters most. It effectively sidesteps the traditional compromise between simplicity and accuracy, offering a smarter, more responsive alternative to static models.
But building a better model is only half the battle. Even the most accurate model will fail to deliver reliable SOC estimates if the estimation algorithm cannot handle real-world noise, disturbances, and sudden changes in driving conditions. To address this, the researchers developed an improved version of the Fractional Extended Kalman Filter (FEKF), a widely used algorithm for state estimation in nonlinear systems.
Standard FEKF algorithms rely on recursive updates that incorporate new measurements over time. However, they can become sluggish or even diverge when faced with abrupt changes—such as sudden acceleration or regenerative braking—because they place too much weight on historical data and not enough on current observations. This inertia can delay the filter’s response and degrade estimation accuracy.
To counteract this, the team introduced a “strong tracking” mechanism by incorporating a time-varying attenuation factor into the filter’s prediction step. This factor dynamically adjusts the state covariance matrix, effectively increasing the influence of recent measurements and reducing the impact of outdated information. The result is a more agile and responsive estimator that can quickly adapt to rapid changes in battery behavior.
The algorithm, named Strong Tracking Fractional Extended Kalman Filter (STF-FEKF), was rigorously tested under three distinct driving cycles: the Urban Dynamometer Driving Schedule (UDDS), the New European Driving Cycle (NEDC), and the Extra-Urban Driving Cycle (EUDC). These profiles simulate a wide range of real-world conditions, from city commuting to highway cruising, allowing for a comprehensive evaluation of the model’s performance.
The results were compelling. Under pulse discharge conditions—a scenario that mimics the sharp current demands of aggressive driving—the FVOM model reduced the average absolute voltage error from 0.0197 V to 0.0160 V, representing an 18.8% improvement in prediction accuracy. More importantly, all voltage errors remained below 50 mV, well within acceptable limits for BMS applications.
When combined with the STF-FEKF estimator, the improvements in SOC estimation were equally impressive. Across all three driving cycles, the new method consistently outperformed the conventional FEKF approach. The average absolute error (AAE) and root mean square error (RMSE) were both significantly reduced, with peak SOC estimation errors staying below 2%, compared to up to 3.2% in the standard method.
Perhaps most telling was the algorithm’s ability to recover quickly from initial estimation errors. In simulations where the starting SOC was set to 80%—a common scenario when a vehicle is plugged in with partial charge—the STF-FEKF algorithm rapidly converged to the true SOC value within minutes, demonstrating robustness and reliability even under suboptimal initialization.
From an engineering perspective, the implications of this research are substantial. For automakers and battery manufacturers, the FVOM model offers a path toward more intelligent and adaptive BMS designs. By enabling more accurate SOC estimation without excessive computational cost, it could lead to longer battery life, improved range prediction, and enhanced safety—all critical factors in consumer adoption of EVs.
Moreover, the use of fractional calculus and adaptive model selection represents a shift toward more physics-informed modeling approaches. Rather than treating the battery as a black box, the FVOM model captures the underlying electrochemical dynamics in a way that is both mathematically rigorous and practically implementable.
The study also highlights the importance of interdisciplinary collaboration in advancing EV technology. Wu Shengli, whose background includes electric vehicle management and signal processing, brought expertise in system modeling and control. Guo Qi contributed to the algorithmic development and experimental validation, while Xing Wenting’s focus on management science and engineering helped frame the problem within broader operational contexts. Together, their combined strengths enabled a holistic approach that bridges theory and application.
Looking ahead, the researchers suggest several directions for future work. One is the extension of the FVOM model to account for temperature variations, which significantly affect battery performance and aging. Another is the integration of aging models to track capacity fade over time, allowing the BMS to adjust its SOC estimates as the battery degrades. Additionally, the team is exploring the potential for real-time implementation on embedded hardware, paving the way for deployment in commercial vehicles.
Independent experts in the field have praised the study for its methodological rigor and practical relevance. “What sets this work apart is its balance between innovation and applicability,” said a battery systems engineer at a leading EV manufacturer who reviewed the paper. “They’re not just proposing a theoretical model—they’ve validated it under realistic conditions and shown measurable improvements. That’s exactly what the industry needs.”
The publication of this research in Energy Storage Science and Technology, a peer-reviewed journal known for its focus on applied energy technologies, underscores its significance within the scientific community. With growing global investment in electrified transportation and grid-scale energy storage, advances in battery modeling are more important than ever.
As governments push for zero-emission transportation and consumers demand greater range and reliability, the pressure on battery technology will only intensify. Solutions like the FVOM model represent a critical step forward—not through revolutionary materials or chemistry, but through smarter, more adaptive software that unlocks the full potential of existing hardware.
In essence, Wu Shengli, Guo Qi, and Xing Wenting have demonstrated that sometimes, the most impactful innovations come not from reinventing the battery, but from rethinking how we understand and manage it. Their work exemplifies how advanced mathematical modeling, when grounded in real-world data and engineering constraints, can drive meaningful progress in sustainable mobility.
For drivers, the benefits may be invisible—no dashboard lights, no alerts—just a more trustworthy, longer-lasting battery that performs as expected, mile after mile. And in the world of electric vehicles, where range anxiety and battery degradation remain top concerns, that kind of quiet confidence is invaluable.
The road to better battery management is not a single leap, but a series of careful, evidence-based steps. This study, with its elegant blend of fractional calculus, statistical model selection, and adaptive filtering, marks a significant stride forward—one that could soon find its way into the next generation of electric vehicles.
Wu Shengli, Guo Qi, Xing Wenting, Chongqing Jiaotong University, Chongqing Technology and Business University, Energy Storage Science and Technology, doi: 10.19799/j.cnki.2095-4239.2024.0174