Lithium battery capacity calculation method

Abstract

The invention relates to the technical field of lithium battery capacity calculation, in particular to a lithium battery capacity calculation method, which comprises the following steps of primary SOC calculation, three-dimensional temperature compensation, curve feature extraction and matching, Kalman filtering fusion and adaptive calibration, breakthrough of voltage platform limitation, elimination of accumulated errors, construction of a full-interval dynamic calibration system and a real-time calibration mechanism. Multi-dimensional coupling compensation, adaptation to an extreme environment and a full life cycle, dynamic curve matching, accurate capture of charge and discharge characteristic changes, minute-level error convergence and meeting of real-time performance and fast charge requirements are realized through five-stage closed loop of primary calculation, dynamic compensation, curve matching, filtering fusion and self-adaptive calibration, time sequence alignment and fast charge are realized, and the real-time performance and the fast charge performance are improved. The DTW algorithm solves curve deviation caused by charge and discharge rate difference, accurately positions SOC nodes, corrects multi-dimensional errors, corrects capacity drift through temperature compensation, suppresses sensor noise through Kalman filtering, and dynamically calibrates and eliminates process errors.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium battery capacity calculation, and specifically to a lithium battery capacity calculation method. Background Art

[0002] As is well known, as a core energy storage component, the accurate calculation of the state of charge (SOC) of a lithium battery is a key technology for the battery management system (BMS). Currently, the mainstream methods include the voltage method: estimating the capacity through the voltage-SOC curve, but for batteries with a flat discharge platform such as lithium iron phosphate, the accuracy in the 20%-80% interval is only above ±5%. The coulomb meter method: calculating the SOC based on current integration, with significant long-term error accumulation. Traditional solutions rely on endpoint calibration, correcting when fully charged / discharged, and the error in the intermediate interval reaches ±3%-5%, the same as above.

[0003] The conventional calculation of lithium battery capacity is the voltage-corresponding charge calculation method and the coulomb meter-corresponding charge. For the voltage-corresponding charge calculation method, especially for lithium batteries with a concentrated discharge platform, for lithium batteries with a flat discharge platform such as lithium iron phosphate, the voltage change amplitude in the 20%-80% SOC interval is less than 0.3V, accounting for about 20% of the total voltage change in the interval, resulting in a fuzzy mapping relationship between voltage and SOC, and the estimation error is as high as above ±5%. The electrochemical characteristics of lithium batteries determine that in the medium capacity interval, such as 20%-80%, it is in a stable electrochemical reaction stage, the ion deintercalation / insertion process of the positive and negative electrode materials is gentle, and the terminal voltage is not sensitive to capacity changes. The traditional voltage method only relies on single-point or linear mapping and cannot capture this non-linear and low-sensitivity change characteristic.

[0004] The coulomb meter-corresponding charge method is relatively accurate, but due to the accuracy problem of the coulomb meter, long-term use will accumulate accuracy errors, resulting in a huge difference between the finally calculated charge and the actual charge. The initial accuracy of the coulomb meter method depends on the current sensor, usually ±0.1% full scale (FS). However, during long-term use, factors such as sensor noise, zero drift, and battery self-discharge will cause the integration error to accumulate continuously, and the error growth rate is about 0.5%-1% per day. The difference between the finally calculated capacity and the actual capacity can reach above ±3%. The current sensor itself has systematic errors, such as temperature drift and long-term aging, and the traditional method does not perform dynamic calibration on the sensor error. Calibration is only triggered when the battery is fully charged, SOC = 100% or discharged, SOC = 0%, relying on the user to actively place the battery in an extreme state. However, in actual applications, such as electric vehicles and energy storage power stations, the battery is rarely fully charged or discharged, resulting in the inability to correct the error in the commonly used 20%-80% interval, forming a calibration blind area. Summary of the Invention

[0005] Technical Problems to be Solved

[0006] In order to overcome the problem that the accuracy of the coulomb meter in the existing lithium battery capacity calculation method will accumulate errors after long-term use, resulting in a huge difference between the finally calculated battery power and the actual battery power, the present invention provides a lithium battery capacity calculation method based on the coulomb meter method and introducing charge-discharge curve matching calibration.

[0007] Technical solution

[0008] To achieve the above object, the present invention provides the following technical solution: a lithium battery capacity calculation method, comprising the following steps:

[0009] Primary SOC calculation: Based on the current data collected in real time by the current sensor, calculate the primary SOC by the coulomb counting method, and the formula is: where Qeff is the effective battery capacity after temperature compensation;

[0010] Three-dimensional temperature compensation: Utilize the pre-calibrated capacity-temperature-aging three-dimensional look-up table Q(T, SOC, Cycle), and dynamically adjust Qeff according to the real-time temperature T, the current SOC, and the battery aging cycle Cycle;

[0011] Curve feature extraction and matching: Sample the voltage curve at 1 Hz in the constant current charge and discharge stages, extract the voltage change rate dV / dt and the current temperature coefficient dI / dT, align the measured curve with the reference curves classified by temperature intervals through the DTW algorithm, and correct the coulomb counting drift at each 10% SOC node;

[0012] Kalman filter fusion: Adopt the extended Kalman filter to perform weighted fusion on the initial value of the coulomb count and the curve-matched SOC, and the observation equation is: V bs = OCV(S|O|C)+I·R i +K·d(V ref -V meas ), to achieve rapid error convergence;

[0013] Adaptive calibration: Trigger calibration in real time in the 20%-80% SOC interval, dynamically adjust the proportional coefficient Kp = 0.5·exp(-0.01∣T - 25∣) and the integral coefficient Ki according to the real-time temperature, and the calibration period is 5 seconds.

[0014] Preferably, the three-dimensional look-up table is generated by performing a cubic polynomial fitting on the battery capacities of -20°C to 60°C, 0%-100% SOC, and 0-500 aging cycles. The reference curves are classified by temperature intervals of -20°C to 0°C, 0°C to 25°C, and 25°C to 60°C, and include the voltage-time curves of constant current charging, constant voltage charging, and constant current discharging modes. The DTW algorithm uses the Euclidean distance as the similarity metric, and finds the optimal time alignment path between the measured curve and the reference curve through dynamic programming, and the alignment error threshold is ±0.05V.

[0015] Furthermore, where w(t) is the process noise, and the covariance matrix is dynamically updated according to the real-time temperature and current fluctuations.

[0016] Furthermore, the calculation of the voltage change rate dV / dt uses a 10-point sliding window filter to suppress high-frequency noise.

[0017] In a further solution, the temperature compensation function of the integral coefficient Ki is: Ki = 0.1·exp(-0.005·∣T - 25∣).

[0018] Based on the foregoing solution, the error convergence determination condition is: the difference in SOC after three consecutive calibrations is less than 0.5% and the voltage residual is less than 0.1V, triggering the adaptive adjustment of the filter gain matrix.

[0019] Based on the foregoing solution, further, the sampling accuracy of the current sensor is ±0.1% FS, the accuracy of the voltage sensor is ±0.2% FS, and the accuracy of the temperature sensor is ±1°C.

[0020] Based on the foregoing solution, further, the reference curve for each temperature range contains sub-curves for different aging cycles (0 - 500 times), and the capacity function Q(T, SOC, Cycle) is generated by cubic polynomial fitting.

[0021] Based on the foregoing solution, further, when the DTW algorithm fails to match, it automatically switches to the reference curve of the adjacent temperature range, and the switching threshold is the Euclidean distance of 0.1V.

[0022] Based on the foregoing solution, further, the multi-stage correction algorithm is integrated into the battery management system microcontroller, and the calculation period is 100ms, meeting the real-time requirement.

[0023] Beneficial effects

[0024] This lithium battery capacity calculation method:

[0025] 1. Break through the voltage platform limit and achieve high-precision estimation in the common range:

[0026] Provide real-time initial values based on coulomb counting method, avoid relying on a single voltage mapping, dynamically match charge and discharge curves through DTW algorithm, calibrate the voltage change rate characteristics (dV / dt) at each 10% SOC node (such as 30%, 50%, 70%), capture the curve shape differences rather than single-point voltage values, reduce the error in the 20%-80% interval from ±5.1% to ±2.7%, with a 47.1% improvement (measured data), achieve precise calibration in the medium-capacity range for the first time, solve the "voltage blind area" problem caused by the flat discharge platform of batteries such as lithium iron phosphate, and provide a more reliable capacity benchmark for the battery management system.

[0027] 2. Eliminate cumulative errors and construct a full-range dynamic calibration system:

[0028] Real-time calibration mechanism: Trigger calibration once every 5 seconds in the 20%-80% interval, without relying on full charge / discharge of the battery, actively correct the current integration drift, Kalman filter fusion: Weightedly fuse the initial value of the coulomb meter and the curve-matched SOC, dynamically suppress sensor noise through the observation equation (including voltage residual and internal resistance voltage drop), and form a closed-loop error correction, Full-range calibration coverage: Expand from the traditional method that can only calibrate at the endpoints of 0-20% and 80-100% to full-range dynamic correction. The measured error in the 0-20% range is reduced from ±3.2% to ±1.5% (a 53.1% improvement), and the error in the 80-100% range is reduced from ±2.8% to ±1.2% (a 57.1% improvement), Cumulative error control: Through high-frequency calibration (once every 5 seconds) and filtering algorithms, suppress the growth rate of long-term use error below ±0.1% / day, and significantly extend the calibration-free cycle of the battery management system.

[0029] 3. Multidimensional coupling compensation to adapt to extreme environments and the full life cycle:

[0030] Three-dimensional look-up table modeling: Pre-calibrate the capacity relationship of temperature (-20°C to 60°C), SOC (0%-100%), and aging cycle (0-500 times), generate the dynamic effective capacity Qeff through cubic polynomial fitting, and replace the fixed rated capacity, Temperature-sensitive parameter self-adaptation: The proportional coefficient Kp = 0.5·exp(-0.01∣T - 25∣) is non-linearly adjusted with temperature, increasing the compensation weight at low temperatures and suppressing noise interference at high temperatures, Extreme temperature adaptability: The capacity calculation accuracy is improved by 60% at -20°C and the error is reduced by 40% at 60°C, effectively solving the non-linear problem of capacity decay caused by temperature, Aging decay compensation: Update the look-up table every 50 cycles, and real-time track the capacity decay (such as when the capacity decays by 15% after 500 cycles, the estimation error is still <±2%), with a 70% reduction in error compared to the traditional fixed-capacity model.

[0031] 4. Dynamic curve matching to accurately capture the changes in charge and discharge characteristics:

[0032] DTW (Dynamic Time Warping): Using Euclidean distance as the metric, it aligns the measured curve with the reference curve through dynamic programming, allowing local stretching / compression of the time axis (such as time alignment between 1C charging curve and 0.5C charging curve), with the matching error threshold controlled within ±0.05V. Feature parameter extraction: Collect the voltage change rate \(dV / dt\) and current temperature coefficient \(dI / dT\) as curve features, replacing the comparison of a single voltage point to capture dynamic information such as curve slope and inflection points. Adaptability to complex working conditions: At different charge and discharge rates (0.5C - 2C) and temperature fluctuations (±10°C), the curve matching accuracy is increased to over 95%, significantly optimized compared to the traditional threshold method (accuracy 70%). Anti-drift ability: When the curve is globally shifted due to battery aging (such as the voltage platform decreasing by 0.1V), the DTW algorithm can still correctly locate the SOC node through shape matching, avoiding mis-calibration.

[0033] 5. Minute-level error convergence, meeting real-time and fast charging requirements:

[0034] Kalman filter for fast fusion: The state equation predicts the SOC change in real time, and the observation equation integrates the voltage residual and curve matching correction amount to achieve 100ms-level update through recursive calculation. High-frequency calibration mechanism: Calibration is triggered every 5 seconds in the 20% - 80% interval, combined with a 1Hz curve sampling rate to capture characteristic changes during charge and discharge in real time. Fast error convergence: The measured error converges to below ±1% within 1 minute, improving the efficiency by over 90% compared to traditional methods (requiring a 4 - 6-hour complete cycle). Adaptability to dynamic scenarios: In the fast charging scenario of electric vehicles (charging 60% SOC in 30 minutes), the error accumulation is reduced from ±3% of traditional methods to ±0.8%, significantly improving charging safety and the reliability of remaining battery life estimation. Brief Description of the Drawings

[0035] Figure 1 It is the flowchart of battery pack power calibration of the present invention;

[0036] Figure 2 It is the schematic diagram of voltage-time curve at different charge and discharge rates of the present invention;

[0037] Figure 3 It is the schematic diagram of voltage-capacity (SOC) relationship curve at different temperatures of the present invention. Detailed Description of the Invention

[0038] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0039] Example 1

[0040] Refer to Figures 1 to 3 , a method for calculating the capacity of a lithium battery, comprising the following steps:

[0041] Primary SOC calculation: Based on the current data collected in real time by the current sensor, the primary SOC is calculated by the Coulomb counting method, and the formula is: where Qeff is the effective battery capacity after temperature compensation;

[0042] Three-dimensional temperature compensation: Using a pre-calibrated capacity-temperature-aging three-dimensional look-up table Q(T, SOC, Cycle), Qeff is dynamically adjusted according to the real-time temperature T, the current SOC, and the battery aging cycle Cycle;

[0043] Curve feature extraction and matching: Sample the voltage curve at 1 Hz during the constant current charge and discharge stages, extract the voltage change rate dV / dt and the current temperature coefficient dI / dT, align the measured curve with the reference curves classified by temperature intervals through the DTW algorithm, and correct the Coulomb counting drift at each 10% SOC node;

[0044] Kalman filter fusion: Use the extended Kalman filter to perform weighted fusion on the initial value of Coulomb counting and the curve-matched SOC, and the observation equation is: V bs = OCV(S|O|C)+I·R i +K·d(V ref -V meas ), to achieve rapid error convergence;

[0045] Adaptive calibration: Trigger calibration in real time within the 20%-80% SOC range, dynamically adjust the proportionality coefficient Kp = 0.5·exp(-0.01∣T - 25∣) and the integral coefficient Ki according to the real-time temperature, and the calibration period is 5 seconds.

[0046] First, in this embodiment, the three-dimensional look-up table is generated by performing a cubic polynomial fitting on the battery capacities at -20°C to 60°C, 0%-100% SOC, and 0-500 aging cycles. The reference curves are classified by temperature intervals of -20°C to 0°C, 0°C to 25°C, and 25°C to 60°C, and include voltage-time curves in constant current charging, constant voltage charging, and constant current discharging modes. The DTW algorithm uses the Euclidean distance as the similarity metric and finds the optimal time alignment path between the measured curve and the reference curve through dynamic programming, and the alignment error threshold is ±0.05V.

[0047] Then, in this embodiment, where w(t) is the process noise, and the covariance matrix is dynamically updated according to the real-time temperature and current fluctuations.

[0048] Secondly, in this embodiment, the calculation of the voltage change rate dV / dt adopts a 10-point sliding window filter to suppress high-frequency noise.

[0049] Thirdly, in this embodiment, the temperature compensation function of the integral coefficient Ki is: Ki = 0.1·exp(-0.005·∣T - 25∣).

[0050] In addition, in this embodiment, the error convergence determination condition is: the difference in SOC after three consecutive calibrations is less than 0.5% and the voltage residual is less than 0.1V, triggering the adaptive adjustment of the filter gain matrix. The sampling accuracy of the current sensor is ±0.1% FS, the voltage sensor accuracy is ±0.2% FS, and the temperature sensor accuracy is ±1°C. The reference curve for each temperature range contains sub-curves for different aging cycles (0 - 500 times), and the capacity function Q(T, SOC, Cycle) is generated by cubic polynomial fitting.

[0051] Finally, in this embodiment, when the DTW algorithm fails to match, it automatically switches to the reference curve of the adjacent temperature range, and the switching threshold is the Euclidean distance of 0.1V. The multi-stage calibration algorithm is integrated into the battery management system microcontroller, and the calculation period is 100ms, meeting the real-time requirement.

[0052] Embodiment 2

[0053] Calculation of the capacity of a 100Ah lithium iron phosphate battery (25°C, new battery)

[0054] Initial configuration

[0055] Sensors: current sensor (200A range, ±0.1% FS), voltage sensor (50V range, ±0.2% FS);

[0056] Reference curve: constant current charging curve in the 25°C range (1C, cutoff at 3.65V), pre-stored voltage-time series.

[0057] Step 1: Primary SOC calculation (t = 0s, SOC = 50%), charging current I = 10A. After 10 seconds, the primary primary SOC: SOC initial value (10) = 50% - 100Ah×3600s / h÷(10A×10s)≈49.972%.

[0058] Step 2: Temperature compensation (T = 25°C, Cycle = 0), Qeff = 100Ah, temperature compensation coefficient Kp = 0.5, Ki = 0.1.

[0059] Step 3: Curve matching (t = 60s, constant current charging stage)

[0060] Collect the voltage sequence, calculate dV / dt = 0.001 V / s, align it with the reference curve at 25°C through DTW, and find that the measured voltage at the 50% SOC node is 0.02 V lower. Correct the Coulomb count: SOC correction = 49.972% + 0.028% = 50.0%.

[0061] Step 4: Kalman filter fusion

[0062] State prediction: SOC(70) ≈ 49.944, observation input: Vmeas = 3.45 V, Vref = 3.47 V, residual 0.02 V, fusion output: SOC fusion = 50.0% (error converges to 0).

[0063] Step 5: Adaptive calibration (continuous charging)

[0064] Trigger calibration every 5 seconds, dynamically adjust Kp and Ki to ensure that the error in the 20% - 80% interval is < ±1.5%.

[0065] Working principle:

[0066] This lithium battery capacity calculation method realizes through a five-stage closed-loop of "primary calculation - dynamic compensation - curve matching - filtering fusion - adaptive calibration":

[0067] Time series alignment: The DTW algorithm solves the curve offset caused by the difference in charge and discharge rates and accurately locates the SOC node.

[0068] Multi-dimensional error correction: Temperature compensation corrects capacity drift, Kalman filter suppresses sensor noise, and dynamic calibration eliminates process errors.

[0069] Full life cycle adaptation: The three-dimensional look-up table is updated with the aging cycle to maintain long-term use accuracy.

[0070] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for calculating the capacity of a lithium battery, characterized in that, It includes the following steps: Primary SOC calculation: Based on the current data collected in real time by the current sensor, the primary SOC is calculated by the Coulomb counting method, and the formula is: where Qeff is the effective battery capacity after temperature compensation; Three-dimensional temperature compensation: Using a pre-calibrated three-dimensional look-up table of capacity-temperature-aging Q(T, SOC, Cycle), dynamically adjust Qeff according to the real-time temperature T, the current SOC, and the battery aging cycle Cycle; Curve feature extraction and matching: Sample the voltage curve at 1 Hz during the constant current charge and discharge stages, extract the voltage change rate dV / dt and the current temperature coefficient dI / dT, align the measured curve with the reference curves classified by temperature intervals through the DTW algorithm, and correct the Coulomb count drift at each 10% SOC node; Kalman filter fusion: The extended Kalman filter is used to perform weighted fusion on the initial value of Coulomb counting and the curve-matched SOC. The observation equation is: V bs = OCV(S│O│C) + I·R i + K·d(V ref - V meas ), achieving rapid error convergence; Adaptive calibration: Trigger calibration in real time within the 20%-80% SOC interval, dynamically adjust the proportional coefficient Kp = 0.5·exp(-0.01|T - 25|) and the integral coefficient Ki according to the real-time temperature, and the calibration period is 5 seconds.

2. The method for calculating the capacity of a lithium battery according to claim 1, wherein The three-dimensional look-up table is generated by performing a cubic polynomial fit on the battery capacity at -20°C to 60°C, 0%-100% SOC, and 0 - 500 aging cycles. The reference curves are classified by temperature intervals of -20°C to 0°C, 0°C to 25°C, and 25°C to 60°C, and include voltage-time curves in the constant current charge, constant voltage charge, and constant current discharge modes. The DTW algorithm uses the Euclidean distance as the similarity metric and finds the optimal time alignment path between the measured curve and the reference curve through dynamic programming, with an alignment error threshold of ±0.05 V.

3. The method for calculating the capacity of a lithium battery according to claim 2, wherein where w(t) is the process noise, and the covariance matrix is dynamically updated according to the real-time temperature and current fluctuations.

4. The method for calculating the capacity of a lithium battery according to claim 3, characterized in that, The calculation of the voltage change rate dV / dt uses a 10-point sliding window filter to suppress high-frequency noise.

5. The method for calculating the capacity of a lithium battery according to claim 4, wherein, The temperature compensation function of the integral coefficient Ki is: Ki = 0.1·exp(-0.005·|T - 25|).

6. The method for calculating the capacity of a lithium battery according to claim 5, wherein The error convergence determination condition is: the SOC difference after 3 consecutive calibrations is less than 0.5% and the voltage residual is less than 0.1 V, triggering the adaptive adjustment of the filter gain matrix.

7. The method for calculating the capacity of a lithium battery according to claim 6, wherein The sampling accuracy of the current sensor is ±0.1% FS, the voltage sensor accuracy is ±0.2% FS, and the temperature sensor accuracy is ±1°C.

8. The method for calculating the capacity of a lithium battery according to claim 7, wherein The reference curves for each temperature interval include sub-curves with different aging cycles (0 - 500 times), and the capacity function Q(T, SOC, Cycle) is generated through cubic polynomial fitting.

9. The method for calculating the capacity of a lithium battery according to claim 8, wherein When the DTW algorithm fails to match, it automatically switches to the reference curve in the adjacent temperature interval, and the switching threshold is the Euclidean distance of 0.1 V.

10. The method for calculating the capacity of a lithium battery according to claim 9, wherein The multi-stage correction algorithm is integrated into the battery management system microcontroller, with a calculation period of 100 ms, meeting the real-time requirements.

Images