Introduction: SOC Measurement – The Core Proposition of Lithium Battery Management
State of Charge (SOC), as the core quantitative indicator of the remaining capacity of lithium batteries, directly determines the safety control accuracy, range prediction reliability, and cycle life of the Battery Management System (BMS). Its essence is to infer the available state of residual charge inside the battery through measurable parameters such as voltage, current, and temperature. However, this process has become an industrial technical challenge due to the strong nonlinear electrochemical characteristics, aging attenuation, temperature interference and other factors of lithium batteries.
In engineering applications, the Ampere-hour Integration Method and Coulomb Counting Method have become the mainstream choices for BMS of small and medium-sized manufacturers due to their simple principles and easy implementation. Research shows that BMS boards of brands such as JK, PACEEX, and DL all adopt the Ampere-hour Integration Method, while JBD prefers the Coulomb Counting Method. But these two methods are not the entire story of SOC measurement, and their inherent limitations have promoted the continuous development of more accurate and robust measurement technologies. This article will systematically sort out the mainstream SOC measurement methods, focus on analyzing the core technologies other than the Ampere-hour Integration Method and Coulomb Counting Method, and build a complete cognitive system of SOC measurement technologies combined with application scenarios and manufacturers’ selection logic.
Basic Groundwork: Two Mainstream Applied Measurement Methods (Current Status and Limitations)
Before delving into other measurement methods, it is necessary to clarify the core characteristics, application status and inherent defects of the Ampere-hour Integration Method and Coulomb Counting Method – this is not only the basis for industrial applications, but also the starting point for subsequent technological innovations.
Ampere-hour Integration Method: A Basic Solution Prioritizing Engineering Adaptability
The core principle is based on the law of conservation of charge. It calculates the change in battery capacity by integrating the charge-discharge current over time. The core formula is: SOC(t) = SOC(0) – (1/Cₙ) × ∫₀ᵗ I(τ)dτ (where SOC(0) is the initial state of charge, Cₙ is the rated capacity, and I(τ) is the charge-discharge current).
In terms of application status, brands such as JK, PACEEX, and DL choose this method mainly because of its low computational complexity, low requirement for hardware computing power, which can be adapted to low-cost embedded BMS modules, and its fast response speed, which can meet the real-time monitoring needs in conventional scenarios. However, this method has obvious limitations: the initial SOC error will accumulate continuously, leading to significant deviations after long-term use; it is greatly affected by the accuracy of current sensors and battery capacity attenuation, and regular calibration is required to maintain accuracy.
Coulomb Counting Method: An Optimized Choice for Dynamic Working Conditions
As an optimized derivative of the Ampere-hour Integration Method, the Coulomb Counting Method is also based on the principle of current integration, but it emphasizes the accurate measurement of charge transfer more. It optimizes the measurement accuracy under dynamic working conditions by introducing the coulomb efficiency coefficient. Its core advantage is good dynamic performance, which is suitable for scenarios with frequent current changes (such as portable devices and light electric vehicles). On the premise of accurate initial calibration, the cumulative measurement accuracy is relatively high.
JBD’s choice of this method is probably because its products focus on dynamic load scenarios and have higher requirements for the real-time and continuity of power measurement. However, this method still cannot get rid of the core limitations: small drifts in current measurement and system noise will gradually accumulate errors; it relies on accurate initial SOC values and regular charge-discharge cycle calibration, and cannot independently cope with the capacity attenuation caused by battery aging.
Core Extension: Mainstream SOC Measurement Technologies Beyond Basic Methods
In response to the inherent defects of the Ampere-hour Integration Method and Coulomb Counting Method, the industry has developed a variety of more accurate and anti-interference measurement methods. These methods are either independently applied in high-precision scenarios or integrated with basic methods to form hybrid architectures, becoming the core technical solutions for mid-to-high-end BMS.
Open Circuit Voltage Method (OCV Method): The “Benchmark Ruler” for Static Calibration
The core principle is to use the strong corresponding relationship between the open circuit voltage (voltage between the two poles in a static state) of lithium batteries and SOC, and infer the SOC value through a pre-calibrated OCV-SOC curve. In engineering, a sixth-order polynomial fitting curve is often used, the fitting error can be controlled within 0.027V, the static estimation accuracy is extremely high, and the average error can be less than 0.5%.
The applicable scenario is mainly the SOC calibration link, such as after charging or long-term parking, to correct the cumulative error of the Ampere-hour Integration Method/Coulomb Counting Method – most automobile manufacturers’ BMS will automatically start the OCV method to calibrate the initial SOC value when the vehicle is stationary for more than 1 hour. However, its fatal limitation is that the battery needs to be stationary for a long time to eliminate the polarization effect, which cannot adapt to real-time working conditions such as dynamic driving and load fluctuations. Therefore, it is rarely used as an independent main measurement method, and is mostly used as a supplementary means to basic methods.
Kalman Filter Series Algorithms: The “Accuracy Corrector” for Dynamic Scenarios
As the mainstream model-based method, the Kalman filter series algorithms dynamically correct SOC estimation errors through constructing battery equivalent circuit models and combining the “prediction-update” iterative mechanism, adapting to nonlinear and strong-interference real-time working conditions, and occupying more than 80% of the mid-to-high-end automotive BMS market. Its core derivative technologies include:
- Extended Kalman Filter (EKF): Linearizes the nonlinear battery system, with low implementation cost and strong real-time performance, suitable for conventional dynamic working conditions. However, linearization approximation will introduce inherent errors, and the accuracy will decrease under strong dynamic scenarios;
- Unscented Kalman Filter (UKF): Uses unscented transformation to generate sampling points to approximate the Gaussian distribution, without the need for linearization, and can capture the high-order moment information of the system. The accuracy is more than 30% higher than EKF under complex working conditions such as NEDC;
- Square Root Cubature Kalman Filter (SRCKF): Avoids the matrix positive semi-definite problem through Cholesky decomposition, and its stability is significantly better than the traditional Kalman filter, adapting to industrial and automotive scenarios with high reliability requirements.
In terms of application cases, automakers such as Tesla and BYD all adopt the hybrid architecture of “Ampere-hour Integration + EKF/UKF”, combined with OCV calibration, to achieve an estimation accuracy of ±3% or less under dynamic working conditions.
Impedance Method: A “Collaborative Measurement Solution” Combining Health Status
The core principle is to measure the battery internal resistance or impedance spectrum, and realize the collaborative estimation of SOC and State of Health (SOH) by using the correlation between internal resistance and SOC, SOH. The battery internal resistance fluctuates regularly with SOC changes (for example, the internal resistance is larger in the low SOC and high SOC stages, and relatively stable in the middle range), and the SOC value can be inferred through high-frequency impedance measurement.
The advantage of this method is that it can reflect the battery aging state synchronously, provide a dynamic capacity benchmark for SOC measurement, and reduce errors caused by aging; but the limitation is that real-time measurement is complex, greatly affected by temperature and charge-discharge rate, and the hardware implementation cost is high. At present, it is mostly used as an auxiliary measurement method, integrated with the Kalman filter method.
Deep Learning Data-Driven Algorithm: An “Intelligent Solution” for Cutting-Edge Breakthroughs
Based on massive battery operation data (voltage, current, temperature, cycle times, etc.), it learns the nonlinear mapping relationship between SOC and multiple parameters through neural network models, with the Long Short-Term Memory (LSTM) algorithm as a typical representative. Its core advantage is strong adaptability, which can automatically adapt to complex factors such as battery aging and temperature fluctuations, without relying on an accurate battery physical model.
At present, this method is in the stage of laboratory verification to engineering transformation, with two core bottlenecks: first, it relies on massive labeled samples (needing to cover the full temperature range of -20℃ to 60℃ and multi-rate scenarios of 0.2C-3C); second, it lacks physical interpretability, making safety verification difficult. The current mainstream research direction is the “model + data-driven” hybrid architecture, such as using LSTM to correct the model error of Kalman filter, taking into account both accuracy and reliability.
Fractional-Order Model Optimization Algorithm: A “Precision Upgrade Solution” for Detail Optimization
Breaking through the limitations of the traditional integer-order RC model, it introduces fractional-order capacitive elements (CPE) to construct an equivalent circuit model, which can more accurately characterize the memory effect and hysteresis characteristics of the battery. Combined with multi-innovation adaptive technology to dynamically adjust the filter gain, it adapts to non-Gaussian noise environments. Under UDDS urban congestion working conditions, the voltage fitting error is reduced by 40% compared with the second-order RC model, providing a more reliable model basis for SOC measurement.
The core challenge of this method is the high computational complexity of fractional-order derivatives, which needs to adapt to the computing power requirements of automotive-grade microcontrollers. It has not yet been mass-produced and applied, but it has become an important direction for high-end BMS technology pre-research.
Different SOC measurement methods have significant differences in accuracy, complexity, cost and applicable scenarios. Manufacturers’ selection is essentially a balance of “demand-cost-accuracy”. Combined with the aforementioned manufacturer cases such as JK, PACEEX, DL, JBD, and mainstream industrial practices, the following core logic can be summarized:
| Measurement Method | Core Advantages | Core Limitations | Typical Application Manufacturers/Scenarios |
| Ampere-hour Integration Method | Low cost, simple implementation, fast response | Error accumulation, relying on calibration | JK, PACEEX, DL (mid-to-low-end BMS boards) |
| Coulomb Counting Method | Good dynamic performance, accurate cumulative measurement | Relying on initial value, requiring regular calibration | JBD (BMS for dynamic load scenarios) |
| Open Circuit Voltage Method | Extremely high static accuracy | Needing long-term static state, unable to apply dynamically | All manufacturers (SOC calibration link) |
| Kalman Filter Series | High dynamic accuracy, strong anti-interference | High complexity, requiring model calibration | Tesla, BYD (mid-to-high-end automotive BMS) |
| Deep Learning Algorithm | Strong adaptability, suitable for complex factors | Relying on data, difficult verification | Research institutions + leading automakers (pre-research stage) |
Future Trends: Multi-Method Integration and Technological Innovation Directions
With the deepening application of lithium batteries in new energy vehicles, energy storage and other fields, the requirements for SOC measurement accuracy and reliability continue to increase. A single method can no longer meet the full-scenario needs. The core development directions in the future present two major characteristics:
In-Depth Integration of Multiple Methods Becomes Mainstream
Basic methods (Ampere-hour Integration/Coulomb Counting) provide a real-time measurement framework, Kalman filter series algorithms dynamically correct errors, OCV method regularly calibrates benchmarks, and impedance method synchronously monitors SOH to dynamically adjust capacity parameters – this “multi-integration” architecture can achieve high-precision measurement within ±2%, which has become the standard technical route for leading automakers’ BMS. For example, the ASRCKF-EKF combined architecture (Adaptive Square Root Cubature Kalman Filter + Extended Kalman Filter) can control the average error within 0.12%-0.16% under various working conditions such as high-speed cruising and urban congestion.
Collaborative Innovation of Data-Driven and Model-Driven Approaches
Through massive operation data from vehicle terminals and energy storage terminals to train deep learning models, optimize the model parameters and noise suppression strategies of Kalman filters, and use physical models to provide constraints for deep learning to solve the problem of insufficient physical interpretability. In addition, with the improvement of chip computing power, the engineering application of high-precision algorithms such as fractional-order models and multi-filter combination will accelerate, further breaking through the measurement bottleneck under complex working conditions.
Lithium battery SOC measurement is not a single technical path, but a complete system of “basic methods as the foundation, precise methods for optimization, and multi-technical integration for upgrading”. The Ampere-hour Integration Method and Coulomb Counting Method are still the mainstream choices for BMS of small and medium-sized manufacturers due to their engineering adaptability, but limited by defects such as error accumulation, they cannot meet the needs of high-end scenarios; technologies such as the Open Circuit Voltage Method, Kalman filter series algorithms, and Impedance Method make up for the shortcomings of basic methods through their respective advantages, forming the core support for high-precision measurement; cutting-edge technologies such as deep learning and fractional-order models provide possibilities for more accurate and adaptable measurement solutions in the future.
For manufacturers, the selection of SOC measurement methods must be closely linked to product positioning and application scenarios – mid-to-low-end products can give priority to the Ampere-hour Integration Method/Coulomb Counting Method, combined with OCV calibration to control costs; mid-to-high-end products need to adopt multi-method integration architectures to balance accuracy and reliability. In the future, with the continuous iteration of technology, SOC measurement will develop towards the direction of “higher accuracy, stronger robustness, and lower cost”, providing core guarantees for the safe and efficient application of lithium batteries.
