A lithium iron phosphate battery pack dynamic pressure difference suppression method and system

Abstract

The application discloses a kind of lithium iron phosphate battery pack dynamic pressure difference inhibition method and system, comprising the following steps: dynamic sorting;Group matching;SOC correction;Predetermined equalization.The application quantifies the voltage rebound characteristics after discharge cutoff, builds a dynamic polarization consistency sorting model, reduces the initial dynamic pressure difference of group to below 8mV from 15mV of traditional method, with a decrease of more than 46%;Based on neural network, the rebound voltage-SOC mapping relationship is established, and the voltage platform area SOC estimation error is corrected, and the error range is compressed from 8% to ±3%;Combined with real-time rebound voltage difference, the pressure difference trend is predicted, and when ΔV_max>15mV or Δt_max>20s, pre-charging equalization is triggered, the polarization difference accumulation is inhibited from the source, and the dynamic pressure difference increment is only 18% after 300 cycles, which significantly prolongs the service life of battery pack and improves system reliability.

Description

Technical Field

[0001] This invention relates to the field of lithium-ion battery pack management technology, and in particular to a method and system for suppressing dynamic voltage difference in lithium iron phosphate battery packs. Background Technology

[0002] Lithium iron phosphate (LFP) batteries have a flat voltage plateau (3.2-3.3V), which makes it impossible for traditional static sorting methods (capacity / internal resistance) to effectively screen for cell inconsistencies under dynamic operating conditions. The voltage difference continues to accumulate during charging and discharging, accelerating the capacity decay of the battery pack.

[0003] In existing technologies, lithium iron phosphate battery packs lack sufficient dimensions for selection: the voltage rebound characteristics after discharge cutoff are ignored (which are strongly correlated with polarization internal resistance and active material distribution), and dynamic polarization differences cannot be identified;

[0004] SOC estimation error: In the flat voltage plateau region, the traditional SOC estimation error is over 8%, which can cause false triggering of equalization.

[0005] Equilibrium lag: Equilibrium is triggered by voltage threshold and cannot be intervened in the early stage of polarization difference accumulation. The peak dynamic differential pressure increases by 45% after 300 cycles. Summary of the Invention

[0006] The purpose of this invention is to provide a dynamic voltage difference suppression method and system for lithium iron phosphate battery packs to solve the above-mentioned problems.

[0007] According to one aspect of the present invention, a method for suppressing dynamic voltage drop in a lithium iron phosphate battery pack is provided, comprising the following steps:

[0008] (a) Dynamic sorting: Collect the voltage rebound curve of the cell after discharge cutoff from 0 to 30 minutes, and extract the voltage value V_rebound after 5 minutes of rest, the time required to rebound to 90% stable voltage Δt_90%, and the slope of the curve from 0 to 5 minutes;

[0009] (b) Group matching: Select cells that meet the conditions |V_rebound|≤10mV and |Δt_90%|≤15s to form a group;

[0010] (c) SOC correction: Construct a neural network model with V_rebound, Δt_90% and the number of iterations as input parameters, and output the SOC correction value ΔSOC to compensate for the SOC estimation error in the voltage plateau region;

[0011] (d) Predictive equalization: Real-time calculation of the range ΔV_max of V_rebound and the range Δt_max of Δt_90% within the group. When ΔV_max > 15mV or Δt_max > 20s, predictive equalization control is triggered.

[0012] In some implementations, the neural network model in step (c) includes three hidden layer neurons, and the SOC-voltage mapping table is updated when |ΔSOC| > 3%.

[0013] In some implementations, the predictive equalization control specifically involves allocating a higher charging priority to individual cells with higher V_rebound in the next charging cycle.

[0014] A system for suppressing dynamic voltage drop in lithium iron phosphate battery packs includes:

[0015] Voltage acquisition module: configured to acquire the voltage rebound curve of each cell after discharge cutoff;

[0016] Sorting control module: performs dynamic sorting and group matching in steps (a)-(b);

[0017] SOC correction module: Integrates the neural network model from step (c) and outputs ΔSOC;

[0018] Balanced execution module: responds to the pre-judgment conditions of step (d) and performs charging priority scheduling.

[0019] Compared with the prior art, the beneficial effects of this application are as follows:

[0020] This invention constructs a dynamic polarization consistency sorting model by quantifying the voltage rebound characteristics after discharge cutoff (V_rebound, Δt_90%, and curve slope), reducing the initial dynamic voltage difference of the battery pack from 15mV to below 8mV using traditional methods, a reduction of over 46%. Based on a neural network, a rebound voltage-SOC mapping relationship is established to correct the SOC estimation error in the voltage plateau region, compressing the error range from 8% to ±3%. Combining real-time rebound voltage differences with the prediction of voltage difference trends, pre-charge equalization is triggered when ΔV_max > 15mV or Δt_max > 20s, suppressing the accumulation of polarization differences from the source. This results in a dynamic voltage difference increase of only 18% after 300 cycles (compared to 45% in traditional solutions), significantly extending battery pack life and improving system reliability. Attached Figure Description

[0021] Figure 1 This is a voltage rebound curve diagram of the present invention;

[0022] Figure 2 This is a graph showing the plateau region error data of the traditional voltage compensation method of the present invention;

[0023] Figure 3 This is the dynamic balancing execution diagram of the present invention;

[0024] Figure 4 This is a flowchart of the method of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] refer to Figures 1 to 4 This application provides a method for suppressing dynamic voltage drop in lithium iron phosphate battery packs, comprising the following steps:

[0027] (a) Dynamic sorting: Collect the voltage rebound curve of the cell after discharge cutoff from 0 to 30 minutes, and extract the voltage value V_rebound after 5 minutes of rest, the time required to rebound to 90% stable voltage Δt_90%, and the slope of the curve from 0 to 5 minutes;

[0028] (b) Group matching: Select cells that meet the conditions |V_rebound|≤10mV and |Δt_90%|≤15s to form a group;

[0029] (c) SOC correction: Construct a neural network model with V_rebound, Δt_90% and the number of iterations as input parameters, and output the SOC correction value ΔSOC to compensate for the SOC estimation error in the voltage plateau region;

[0030] (d) Predictive equalization: Real-time calculation of the range ΔV_max of V_rebound and the range Δt_max of Δt_90% within the group. When ΔV_max > 15mV or Δt_max > 20s, predictive equalization control is triggered.

[0031] In some implementations, the neural network model in step (c) includes three hidden layer neurons, and the SOC-voltage mapping table is updated when |ΔSOC| > 3%.

[0032] In some implementations, the predictive equalization control specifically involves allocating a higher charging priority to individual cells with higher V_rebound in the next charging cycle.

[0033] A system for suppressing dynamic voltage drop in lithium iron phosphate battery packs includes:

[0034] Voltage acquisition module: configured to acquire the voltage rebound curve of each cell after discharge cutoff;

[0035] Sorting control module: performs dynamic sorting and group matching in steps (a)-(b);

[0036] SOC correction module: Integrates the neural network model from step (c) and outputs ΔSOC;

[0037] Balanced execution module: responds to the pre-judgment conditions of step (d) and performs charging priority scheduling.

[0038] Dynamic sorting model based on rebound voltage:

[0039] Under standard operating conditions (25℃), by quantifying the amplitude, stabilization time, and curve shape of the rebound voltage after discharge cutoff, a dynamic polarization consistency evaluation index for battery cells is constructed to improve the accuracy of group matching.

[0040] Rebound voltage-SOC correlated calibration:

[0041] Establish a mapping relationship between rebound voltage characteristic parameters (5-minute rebound voltage value) and SOC to correct the estimation error of the plateau region of the traditional voltage-SOC curve.

[0042] Predictive equilibrium triggering mechanism:

[0043] By combining the rebound voltage difference to predict the dynamic differential pressure trend, equilibration is initiated before the differential pressure significantly expands, thus avoiding a delayed response.

[0044] refer to Figure 1 In step (a), the test conditions are as follows: In a constant temperature environment of 25℃, the cells to be sorted are discharged at a 1C rate to the cutoff voltage (2.0V), left to stand, and the voltage rebound curve for 0-30 minutes after cutoff is recorded. Feature extraction: Key parameters of the rebound voltage are calculated, including:

[0045] V_rebound: Voltage value after 5 minutes of rest;

[0046] Δt_90%: The time required for the voltage to rebound to 90% of its stable value;

[0047] Curve slope: Voltage change rate within 0-5 minutes.

[0048] Dynamic sorting: Cells with V_rebound difference ≤10mV and Δt_90% difference ≤15s are selected by clustering algorithm and grouped together (grouping is done by comparing the dynamic differences of each individual cell).

[0049] Rebound voltage-SOC correlation modeling

[0050] At 25℃, multiple charge-discharge cycles were performed on the battery cells, and the rebound voltage parameters after discharge cutoff at different SOC points (20%, 50%, 80%) were recorded.

[0051]

[0052]

[0053]

[0054]

[0055] A neural network model is constructed, with input parameters including V_rebound, Δt_90%, and the number of historical charge-discharge cycles. The output is the SOC correction value ΔSOC, which compensates for the plateau error of the traditional voltage method. Figure 2 Verification logic: If ΔSOC > 3%, trigger the cell capacity calibration mode and update the SOC-voltage mapping table.

[0056] SOC-Rebound Voltage Correction Model Architecture: Input layer (V_rebound, Δt_90%, number of iterations) → Hidden layer (3 layers of neurons) → Output layer (ΔSOC).

[0057] Predictive Equilibrium Control

[0058] Real-time monitoring: After the battery pack is discharged, collect the V_rebound and Δt_90% of each individual cell.

[0059] Voltage differential trend prediction: After discharge cutoff, the ranges ΔV_max and Δt_90% of the cell rebound within the group are calculated in real time. If ΔV_max > 15mV or Δt_max > 20s, it is determined to be voltage differential risk level I, and a pre-equilibrium command is generated.

[0060] Dynamic balancing execution: In the next charging cycle, high-V_rebound cells (low polarization resistance) are prioritized for charging to suppress voltage drop accumulation. (Reference) Figure 3 .

[0061] Technical problems to be solved:

[0062] 1. Traditional sorting methods cannot identify differences in dynamic polarization characteristics;

[0063] 2. Low SOC estimation accuracy in the lithium iron phosphate battery platform region leads to false triggering of equalization;

[0064] 3. Lack of early warning in the early stage of dynamic pressure difference formation leads to equilibrium lag.

[0065] By standardizing the acquisition of rebound voltage characteristics after discharge cutoff, the traditional sorting and equalization control is expanded from "real-time voltage monitoring" to "polarization recovery capability assessment", suppressing the generation of dynamic voltage difference from the source, and at the same time, combining predictive control to achieve proactive intervention.

[0066] This invention constructs a dynamic polarization consistency sorting model by quantifying the voltage rebound characteristics after discharge cutoff (V_rebound, Δt_90%, and curve slope), reducing the initial dynamic voltage difference of the battery pack from 15mV to below 8mV using traditional methods, a reduction of over 46%. Based on a neural network, a rebound voltage-SOC mapping relationship is established to correct the SOC estimation error in the voltage plateau region, compressing the error range from 8% to ±3%. Combining real-time rebound voltage differences with the prediction of voltage difference trends, pre-charge equalization is triggered when ΔV_max > 15mV or Δt_max > 20s, suppressing the accumulation of polarization differences from the source. This results in a dynamic voltage difference increase of only 18% after 300 cycles (compared to 45% in traditional solutions), significantly extending battery pack life and improving system reliability.

[0067] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for suppressing dynamic voltage difference in a lithium iron phosphate battery pack, characterized in that, Includes the following steps: (a) Dynamic sorting: Collect the voltage rebound curve of the cell after discharge cutoff from 0 to 30 minutes, and extract the voltage value V_rebound after 5 minutes of rest, the time required to rebound to 90% stable voltage Δt_90%, and the slope of the 0-5 minute curve; (b) Group matching: Select cells that meet the following conditions for grouping: |V_rebound| difference ≤ 10mV and |Δt_90%| difference ≤ 15s; (c) SOC Correction: Construct a neural network model with V_rebound, Δt_90% and the number of iterations as input parameters, and output the SOC correction value ΔSOC to compensate for the SOC estimation error in the voltage plateau region. The neural network model contains three hidden layer neurons. When |ΔSOC|>3%, the SOC-voltage mapping table is updated. (d) Predictive equalization: Real-time calculation of the range ΔV_max of V_rebound and the range Δt_max of Δt_90% within the group. When ΔV_max > 15mV or Δt_max > 20s, predictive equalization control is triggered.

2. The method for suppressing dynamic voltage difference in lithium iron phosphate battery packs according to claim 1, characterized in that, The predictive equalization control specifically involves allocating higher charging priority to individual cells with higher V_rebound in the next charging cycle.

3. A system based on the dynamic voltage difference suppression method for lithium iron phosphate battery packs according to any one of claims 1-2, characterized in that, include: Voltage acquisition module: configured to acquire the voltage rebound curve of each cell after discharge cutoff; Sorting control module: performs dynamic sorting and group matching in steps (a)-(b); SOC correction module: Integrates the neural network model from step (c) and outputs ΔSOC; Balanced execution module: responds to the pre-judgment conditions of step (d) and performs charging priority scheduling.

Images