Inclusive & Assistive Products
ISDN2001/2002: Second Year Design Project
Independent study
Accuracy Testing of Lithium Battery State of Charge (SOC) Estimation Based on Voltage Method
Artur Zhang
Significance
The accurate estimation of the State of Charge (SOC) in lithium batteries is crucial for ensuring battery safety, optimizing performance, and extending battery lifespan. Among various SOC estimation methods, the Open Circuit Voltage (OCV) method is widely used due to its simplicity and reliability. However, factors such as discharge current, temperature, and battery aging can affect its accuracy. This research investigates the accuracy of voltage-based SOC estimation, identifies key influencing factors, and explores potential improvements to enhance BMS (Battery Management System) performance. Though temperature and discharge current are named during proposal stage, it is later found that those parts are not controllable in any experiment condition I can find and very sophisticated for beginners in electronics like me, so the later focus turned more to focusing on 18650 battery level estimation so to implement to our project.
Refined research topic
-
1. To analyze the variation in SOC estimation accuracy using the voltage method.
-
2. To identify key factors affecting the accuracy of voltage-based SOC estimation.
Research methodology
Charge-Discharge Characteristics:
Study the relationship between voltage and capacity under different discharge rates by testing battery voltage at different times and map the graph with Metlab, use metlab for simulating characteristic between 0 and 500 rounds of charge-discharge cycle.
Voltage-Capacity Curve:
Analyze how SOC correlates with OCV under varying conditions.
Experiment Design
-
Test Conditions:
Charge and discharge one battery from 0 cycle to 50 cycles and simulate to 500 cycles. -
Equipment Selection:
18650 battery pack
Discharge circuit with project materials
Multimeter
Thermo gun (for precautions of any possible dumb mistakes made by electronics newbie like me)
Data acquisition system for
voltage and current logging
Metlab for plotting curve base
on experiment data points
Procedure
-
Fully charge the battery to 100% SOC.
-
Discharge at different C-rates while recording voltage and capacity.
-
Compare estimated SOC (based on OCV) with actual discharge capacity.
Research Results
Pt.1 To analyze the variation in SOC estimation accuracy using the voltage method.
Metlab curve simulation after 27 measurements, measuring interval equal or more than 4 hours (image on the right)
Metlab code in python:
SOC_fit = np.linspace(0, 100, 100)
OCV_fit = poly(SOC_fit)
plt.scatter(SOC, OCV)
plt.plot(SOC_fit, OCV_fit, 'r')
plt.xlabel('SOC (%)')
plt.ylabel('OCV (V)')
plt.legend()
plt.grid()
plt.show()
Analysis
The OCV is positively related to SOC, but in details, it obeys following rules:
• High SOC range (80%~100%): OCV changes significantly with SOC (large slope), suitable for voltage estimation.
• Medium SOC range (20%~80%): OCV changes slowly (small slope), and the voltage method has a large error.
• Low SOC range (0%~20%): OCV drops rapidly again, but excessive discharge should be avoided。
And for 18650s used in our product, The slope of the OCV-SOC curve changes significantly, which is suitable for the voltage method. As compared to Lithium iron phosphate (LFP): The OCV in the middle SOC range is almost flat (3.2~3.3V), and the voltage method has extremely low accuracy and needs to be combined with other methods (such as coulomb counting).
Pt.2 2. To identify key factors affecting the accuracy of voltage-based SOC estimation.
Simulated curve on right side, cure from the right to left: 0 cycles (new battery characteristic), 50 cycles, 300 cycles, 500 cycles.
Analysis
New battery (0~50 cycles)
OCV-SOC characteristics
Curve shape: OCV-SOC relationship is stable and conforms to the standard curve provided by the manufacturer (such as the typical "S" curve of 18650).
High/low SOC range:
OCV ≈ 4.2V (full charge voltage) at 100% SOC;
OCV ≈ 3.0V (cut-off voltage) at 0% SOC.
Mid-SOC platform area: The slope changes significantly (such as 18650 at 30%~80% SOC, OCV changes by about 0.5V).
Mid-term aging (50~300 cycles)
OCV-SOC change trend
Voltage platform sinking:
Under the same SOC, the overall OCV decreases (e.g., 4.2V→4.18V at full charge, the voltage in the mid-SOC region decreases by 10~30mV).
Reason: loss of active materials (reduced efficiency of positive electrode Li+ insertion and extraction) and thickening of SEI film lead to increased polarization.
Slope decay:
The OCV change rate in the high/low SOC range decreases (e.g., the voltage difference of 100%~90% SOC changes from 0.15V→0.12V).
Capacity decay: 5%~15%
Deep aging (300~500 cycles)
OCV-SOC deteriorates significantly
Voltage platform drift:
OCV fluctuation increases in the medium SOC range (such as 40%~60%), and the curve flattens (similar to the LFP battery phenomenon).
Terminal voltage drop:
OCV drops faster at low SOC (<10%), which can easily trigger BMS misjudgment (early discharge termination).
Capacity decay: 15%~30%
Final conclusion
The 18650 is suitable for our project as it is easy to detect with low tech and cheap method, considering our product needs one charging cycle in about 2 months, so the difference in decayed 18650 can be almost ignored in our product’s life cycle.
Final BMS circuit design
After understanding the basic princlple, a pcb for simple control over battery system is designed and manufactured, and was implemented in our prototype. As concluded from research, although there are different factors affecting 18650 SOC estimation, but due to the nature of our product, most can be neglected, and the biggest variable (battery aging) is also acceptable, allowing easier and cheaper way of estimating battery level for more compact and servicable product.