A lithium iron phosphate battery soc estimation and correction method based on a fuzzy EKF-AH algorithm

Abstract

The present application relates to the technical field of lithium battery state estimation, and more particularly to a lithium iron phosphate battery SOC estimation and correction method based on a fuzzy EKF-AH algorithm. Starting with a certain start of an energy storage device, the BMS reads the SOC and SOH at the last shutdown, and according to the standing time, selects the open circuit voltage OCV correction or EKF algorithm correction SOC; the SOC is estimated using the EKF-AH algorithm, the fuzzy rule base associated with the state of charge SOC and the health status SOH is established combined with fuzzy control, and the EKF-AH algorithm weight and measurement noise error are dynamically adjusted; the estimated difference of the EKF algorithm and the ampere-hour integral method is calculated, and if the sum of the estimated difference is greater than the corresponding threshold, an SOH correction warning is issued. The estimation accuracy of the whole life cycle of the lithium iron phosphate battery is improved, and the SOH is also corrected.

Description

Technical Field

[0001] This invention relates to the field of lithium battery state estimation technology, and in particular to a method for estimating and correcting the state of charge (SOC) of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm. Background Technology

[0002] To better address environmental pollution and the energy crisis, the development of energy storage technology has garnered significant attention. Lithium-ion batteries, as one of the main power sources for energy storage systems, play a crucial role in energy storage and management. The state of charge (SOC) of a battery is a key indicator for assessing its remaining energy, directly impacting the performance, energy efficiency, and lifespan prediction of the energy storage system. Improving the energy efficiency and performance of energy storage systems, including the accuracy of SOC estimation, is of great significance for achieving dual-carbon goals and promoting energy transition.

[0003] Currently, mainstream SOC estimation methods include the ampere-hour integration method, the open-circuit voltage method, the Kalman filter method, and machine learning algorithms. The ampere-hour integration method is the most common SOC estimation method, calculating the battery SOC by integrating the current over time. This method is significantly affected by sensor accuracy, cannot eliminate absolute errors caused by incorrect initial SOC values, and suffers from the problem of process errors accumulating over time. The open-circuit voltage method estimates the battery's remaining capacity by measuring the SOC values ​​at different open-circuit voltages, but this method requires the battery to be idle for a long time, making real-time monitoring inconvenient, and it is inaccurate in estimating the OCV plateau period of lithium iron phosphate batteries. Machine learning algorithms, represented by neural networks, rely on a large amount of experimental data, resulting in poor robustness and battery universality. In contrast, the Kalman filter method, due to its high estimation accuracy, good robustness, and ability to eliminate the influence of incorrect initial values, has become a research hotspot for SOC estimation methods.

[0004] The essence of Extended Kalman Filter (EKF) estimation of State of Charge (SOC) is to simulate the polarization characteristics of the battery through an equivalent circuit model, and then use the Open Circuit Value (OCV) and Ampere-Hour Integration (ACV) method to jointly estimate the SOC. Therefore, the EKF algorithm incorporates the advantages and disadvantages of the open-circuit voltage method, and is inaccurate during the battery's plateau period (i.e., when the slope of the SOC-OCV curve is too small), easily leading to errors. Furthermore, traditional Kalman filter algorithms assume that the measurement noise error γ is a fixed value and do not consider the impact of battery life variations on the battery's OCV characteristics, resulting in increasingly larger estimation errors. Summary of the Invention

[0005] To address the limitations of existing technologies, this invention proposes a method for estimating and correcting the state of charge (SOC) of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm.

[0006] This invention uses the EKF-AH algorithm, which combines extended Kalman filtering and ampere-hour integration, to estimate the state of charge (SOC) of lithium iron phosphate (LFP) batteries. A fuzzy rule base is established, relating the current SOC to the current state of charge (SOC) and the state of health (SOH) over the battery's lifespan. The weights of the EKF-AH algorithm and the measurement noise error γ are dynamically adjusted, eliminating the initial error and cumulative error of the ampere-hour integration method. This solves the problem of inaccuracy of EKF during the plateau period and improves the estimation accuracy of LFP batteries throughout their entire lifespan. A method for estimating and correcting the SOC of LFP batteries based on the fuzzy EKF-AH algorithm is designed, including the following steps:

[0007] S1, starting with a specific power-on of the energy storage device, the BMS reads the SOC and SOH values ​​from the last power-off, and records them as SOC. his SOH his ;

[0008] S2, calculate the time difference between the current power-on and the last power-off as the battery rest time ΔT, and determine whether the battery rest time ΔT has reached the preset value T of the battery rest time;

[0009] If the battery resting time ΔT≥T, proceed to step S3;

[0010] If the battery resting time ΔT < T, proceed to step S4;

[0011] S3, use the open-circuit voltage OCV to correct the SOC value and obtain the new SOC. new And calculate SOC his With SOC new The difference ΔSOC=|SOC his -SOC new Let the SOH correction warning threshold for the battery be δ;

[0012] If ΔSOC>δ, determine SOH. his If the estimation is incorrect, proceed to step S9;

[0013] If ΔSOC≤δ, then proceed to step S5;

[0014] S4, use the EKF algorithm to correct the SOC. The correction process lasts for 1 minute, yielding the corrected SOC. new Proceed to step S5;

[0015] S5, Input the corrected SOC new Using the EKF algorithm and the ampere-hour integration method as initial values, the SOC is estimated to obtain the SOC. EKF SOC AHBy combining fuzzy control to establish a fuzzy rule base associated with the state of charge (SOC) and the state of health (SOH), the weights ω of the EKF-AH algorithm and the measurement noise error γ are dynamically adjusted to obtain an accurate estimate of SOC. NEW ;

[0016] S6, calculate the estimation difference e(k) = |SOC| between the EKF algorithm and the ampere-hour integration method. EKF -SOC AH Establish TS fuzzy control rules and set different warning thresholds α based on whether SOC is in a plateau period.

[0017] When 10% < SOC NEW If the SOC is less than 90%, meaning it is in a plateau phase, then the warning threshold α is set to 2%.

[0018] When SOC NEW ≤10%, or SOC NEW If the SOC is ≥90%, meaning it is in a non-plateau period, then the warning threshold α is set to 5%.

[0019] S7 is obtained by summing the differences from the first n estimations. If the sum of the estimated differences e(k) is greater than the corresponding warning threshold α, proceed to step S9; otherwise, proceed to step S8.

[0020] S8, BMS updates the current SOC of energy storage devices, providing an accurate SOC estimate. NEW Assign to SOC new Proceed to step S5;

[0021] S9, BMS determines SOH his An estimate was incorrect; a SOH correction warning has been issued.

[0022] Furthermore, in step S3, the open-circuit voltage OCV is used to correct the SOC value, specifically including the following:

[0023] S3.1, Perform battery OCV-SOC test to obtain the relationship between battery open circuit voltage OCV and remaining capacity SOC under different temperatures and aging conditions;

[0024] S3.2, using MATLAB to fit the polynomial of the functional relationship between open-circuit voltage OCV and SOC.

[0025] OCV=f(SOC=a0+a1SOC 1 +a2SOC 2 +…+a n SoC n ,

[0026] Where a n The coefficients of the nth power term;

[0027] S3.3, Substitute the open-circuit voltage OCV read by the sensor after power-on into the polynomial to obtain SOC. new .

[0028] Furthermore, in step S4, the EKF algorithm is used to correct the SOC calculation, specifically including the following:

[0029] S4.1, Based on the polarization characteristics of lithium batteries, establish the Thevenin equivalent circuit model;

[0030] S4.2, the working equations of the Thevenin equivalent circuit model above are:

[0031]

[0032] In the above formula, OCV is the open-circuit voltage, R0 is the ohmic internal resistance of the battery, C is the polarization capacitor, R1 is the polarization resistor, and U1 represents the voltage across the polarization resistor R1 and the polarization capacitor C. s Where is the battery terminal voltage, and i is the current battery current, positive for charging and negative for discharging;

[0033] S4.3 Discretize the above equations and combine them with extended Kalman filtering to obtain the observation equations and measurement equations as follows.

[0034]

[0035] In the observation and measurement equations, t is the unit sampling interval, and Q... max The current maximum capacity of the battery is related to the state of equilibrium (SOH). Q represents the process noise error matrix, and R represents the measurement noise error matrix.

[0036] The algorithm recursively estimates as follows:

[0037] Prior estimation,

[0038]

[0039] Kalman gain update,

[0040]

[0041] Optimal estimate update,

[0042]

[0043] P is the prior estimate of the observed state at the current moment. k - Let K be the prior error covariance matrix at the current time. k Here is the Kalman gain matrix. P k + Let R be the optimal estimate and the updated error covariance matrix at the current time, Q be the process noise error matrix, and R be the error matrix. k Let U be the measurement noise error matrix at the current moment. s This is the value measured by the voltmeter. This is an estimate of the voltage at the opposite end at the current moment;

[0044] S4.4, HPPC test is performed, and the battery is charged and discharged using a continuous pulse sequence current to obtain the dynamic characteristics of the battery under different SOC values. MATLAB is used to identify the battery parameters offline to obtain R0, R1, and C.

[0045] S4.5, Substituting the previously estimated SOC(k-1) and U1(k-1), we obtain SOC. new =SOC(k).

[0046] Furthermore, in S4, the EKF algorithm is used to correct the SOC, and its measurement noise error matrix R = 0.0003.

[0047] Furthermore, the dynamic adjustment of the EKF-AH algorithm weights ω and measurement noise error γ by fuzzy control in S5 specifically includes the following:

[0048] S5.1, Obtain SOC new SOH his Using the ampere-hour integration method and EKF to evaluate the SOC new SOC was estimated separately AH SOC EKF The discretization formula for the ampere-hour integral method is as follows:

[0049]

[0050] Q max =SOH his *Q 出厂 ,

[0051] In the above formula, t is the unit sampling interval, i is the current battery current (positive for charging, negative for discharging), and Q... max Q represents the battery's current maximum capacity. 出厂 Q represents the battery's maximum rated capacity at the time of manufacture. This capacity decreases as the battery is used. max The SOH level continues to decrease;

[0052] S5.2, in the fuzzy controller, set SOC new SOH hisAs input to the fuzzy controller, the EKF-AH algorithm weights ω and the measurement noise error γ are used as the system output;

[0053] Define three fuzzy subsets of SOC as Bigger(B), Middler(M), and Lower(L);

[0054] The three fuzzy subsets of SOH are defined as Old(O), NOW(M), and New(N);

[0055] Similarly, the fuzzy subsets of ω and γ are defined as Big(B), Mid(M), and Low(L), respectively.

[0056] Based on the membership function type and parameter settings, fuzzy inference rules are formulated through the design of a fuzzy controller;

[0057] S5.3, through the design of a fuzzy controller, combined with the range of the State of Charge (SOC) and the magnitude of the State of Health (SOH), the fuzzy controller is used to achieve dynamic adaptive adjustment of the EKF-AH algorithm weights ω and the measurement noise error γ. Finally, the adjusted measurement noise error γ is passed to the measurement noise error matrix R in the Kalman filter gain coefficient, and the latest EKF-AH algorithm weights ω are used to calculate and obtain the accurate SOC estimate. NEW See the formula below.

[0058] SOC NEW (k)=ωSOC AH (k)+(1-ω)SOC EKF (k).

[0059] Furthermore, the corrected warning threshold δ is set to 10%.

[0060] Compared with the prior art, the advantages of the present invention are as follows:

[0061] 1. The EKF-AH algorithm, which combines extended Kalman filtering and ampere-hour integration, is used to estimate the state of charge (SOC) of lithium iron phosphate batteries. This eliminates the initial value error and cumulative process error of the ampere-hour integration method, solves the problem of inaccuracy of EKF during the plateau period, and improves the accuracy of SOC estimation.

[0062] 2. A fuzzy rule base was established that is associated with the current state of charge (SOC) and health status (lifetime) (SOH). The weights of the EKF-AH algorithm and the measurement noise error γ were dynamically adjusted, which improved the estimation accuracy of the entire life cycle of lithium iron phosphate batteries.

[0063] 3. Set different thresholds based on whether SOC is in a plateau period, compare the cumulative estimation error of the two algorithms with the threshold size, determine whether SOH estimation is wrong, and then assist in SOH correction. Attached Figure Description

[0064] Figure 1 This is a flowchart of the SOC estimation and correction method for lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm in an embodiment of the present invention;

[0065] Figure 2 This is the Thevenin equivalent circuit model in the embodiments of the present invention;

[0066] Figure 3 This is an OCV-SOC curve of the battery under different SOH conditions in the embodiments of the present invention;

[0067] Figure 4 The EKF algorithm estimation effect in this embodiment of the invention;

[0068] Figure 5 The EKF-AH algorithm estimation effect is shown in the embodiment of the present invention. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be further described below. It should be noted that the embodiments described in the following examples do not represent all embodiments consistent with this disclosure.

[0070] A method for estimating and correcting the state of charge (SOC) of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm includes the following steps:

[0071] S1, starting with a specific power-on of the energy storage device, the BMS reads the SOC and SOH values ​​from the last power-off, and records them as SOC. his SOH his ;

[0072] S2, calculate the time difference between the current power-on and the last power-off as the battery rest time ΔT, and determine whether the battery rest time ΔT has reached the preset value T of the battery rest time;

[0073] If the battery resting time ΔT ≥ T, proceed to step S3; if the battery resting time ΔT < T, proceed to step S4.

[0074] S3, use the open-circuit voltage OCV to correct the SOC value and obtain the new SOC. new And calculate SOC his With SOC new The difference ΔSoC = |SOC his -SOC new Let the SOH correction warning threshold for the battery be δ;

[0075] If ΔSOC>δ, determine SOH. his If the estimation is incorrect, proceed to step S9;

[0076] If ΔSOC≤δ, then proceed to step S5;

[0077] S4. The SOC is corrected using the EKF algorithm. The correction process lasts for 5 minutes, yielding the corrected SOC. new Proceed to step S5;

[0078] S5, Input the corrected SOC new Using the EKF algorithm and the ampere-hour integration method as initial values, the SOC is estimated to obtain the SOC. EKF SOC AH By combining fuzzy control to establish a fuzzy rule base associated with the state of charge (SOC) and the state of health (SOH), the weights ω of the EKF-AH algorithm and the measurement noise error γ are dynamically adjusted to obtain an accurate estimate of SOC. NEW ;

[0079] S6, calculate the estimation difference e(k) = |SOC| between the EKF algorithm and the ampere-hour integration method. EKF -SOC AH Establish TS fuzzy control rules and set different warning thresholds α based on whether SOC is in a plateau period.

[0080] When 10% < SOC NEW If the SOC is less than 90%, meaning it is in a plateau phase, then the warning threshold α is set to 2%.

[0081] When SOC NEW ≤10%, or SOC NEW If the SOC is ≥90%, meaning it is in a non-plateau period, then the warning threshold α is set to 5%.

[0082] S7 is obtained by summing the differences from the first n estimations. If the sum of the estimated differences e(k) is greater than the corresponding warning threshold α, proceed to step S9; otherwise, proceed to step S8.

[0083] S8, BMS updates the current SOC of energy storage devices, providing an accurate SOC estimate. NEW Assign to SOC new Proceed to step S5;

[0084] S9, BMS determines SOH his An estimate was incorrect; a SOH correction warning has been issued.

[0085] Furthermore, in step S3, the open-circuit voltage OCV is used to correct the SOC value, specifically including the following:

[0086] S3.1, Perform battery OCV-SOC test to obtain the relationship between battery open circuit voltage OCV and remaining capacity SOC under different temperatures and aging conditions;

[0087] S3.2, using MATLAB to fit the polynomial of the functional relationship between open-circuit voltage OCV and SOC.

[0088] OCV=f(SOC=a0+a1SOC 1 +a2SOC 2 +…+a n SOC n ,

[0089] Where a n The coefficients of the nth power term;

[0090] S3.3, Substitute the open-circuit voltage OCV read by the sensor after power-on into the polynomial to obtain SOC. new .

[0091] Furthermore, in step S4, the EKF algorithm is used to correct the SOC calculation, specifically including the following:

[0092] S4.1, Based on the polarization characteristics of lithium batteries, establish the Thevenin equivalent circuit model;

[0093] S4.2, the working equations of the Thevenin equivalent circuit model above are:

[0094]

[0095] In the above formula, OCV is the open-circuit voltage, R0 is the ohmic internal resistance of the battery, C is the polarization capacitor, R1 is the polarization resistor, and U1 represents the voltage across the polarization resistor R1 and the polarization capacitor C. s Where is the battery terminal voltage, and i is the current battery current, positive for charging and negative for discharging;

[0096] S4.3 Discretize the above equations and combine them with extended Kalman filtering to obtain the observation equations and measurement equations as follows.

[0097]

[0098] In the observation and measurement equations, t is the unit sampling interval, and Q... max The current maximum capacity of the battery is related to the state of equilibrium (SOH). Q represents the process noise error matrix, and R represents the measurement noise error matrix.

[0099] The algorithm recursively estimates as follows:

[0100] Prior estimation,

[0101]

[0102] Kalman gain update,

[0103]

[0104] Optimal estimate update,

[0105]

[0106] P is the prior estimate of the observed state at the current moment. k - Let K be the prior error covariance matrix at the current time. k Here is the Kalman gain matrix. P k + Let R be the optimal estimate and the updated error covariance matrix at the current time, Q be the process noise error matrix, and R be the error matrix. k Let U be the measurement noise error matrix at the current moment. s This is the value measured by the voltmeter. This is an estimate of the voltage at the opposite end at the current moment;

[0107] S4.4, HPPC test is performed, and the battery is charged and discharged using a continuous pulse sequence current to obtain the dynamic characteristics of the battery under different SOC values. MATLAB is used to identify the battery parameters offline to obtain R0, R1, and C.

[0108] S4.5, Substituting the previously estimated SOC(k-1) and U1(k-1), we obtain SOC. new =SOC(k).

[0109] Furthermore, in S4, the EKF algorithm is used to correct the SOC, and its measurement noise error matrix R = 0.0003.

[0110] Furthermore, the dynamic adjustment of the EKF-AH algorithm weights ω and measurement noise error γ by fuzzy control in S5 specifically includes the following:

[0111] S5.1, Obtain SOC new SOH his Using the ampere-hour integration method and EKF to evaluate the SOC new SOC was estimated separately AH SOC EKF The discretization formula for the ampere-hour integral method is as follows:

[0112]

[0113] Q max =SOHhis *Q 出厂 ,

[0114] In the above formula, t is the unit sampling interval, i is the current battery current (positive for charging, negative for discharging), and Q... max Q represents the battery's current maximum capacity. 出厂 Q represents the battery's maximum rated capacity at the time of manufacture. This capacity decreases as the battery is used. max The SOH level continues to decrease;

[0115] S5.2, in the fuzzy controller, set SOC new SOH his As input to the fuzzy controller, the EKF-AH algorithm weights ω and the measurement noise error γ are used as the system output;

[0116] Define three fuzzy subsets of SOC as Bigger(B), Middler(M), and Lower(L);

[0117] The three fuzzy subsets of SOH are defined as Old(O), NOW(M), and New(N);

[0118] Similarly, the fuzzy subsets of ω and γ are defined as Big(B), Mid(M), and Low(L), respectively.

[0119] Based on the membership function type and parameter settings, fuzzy inference rules are formulated through the design of a fuzzy controller;

[0120] S5.3, through the design of a fuzzy controller, combined with the range of the State of Charge (SOC) and the magnitude of the State of Health (SOH), the fuzzy controller is used to achieve dynamic adaptive adjustment of the EKF-AH algorithm weights ω and the measurement noise error γ. Finally, the adjusted measurement noise error γ is passed to the measurement noise error matrix R in the Kalman filter gain coefficient, and the latest EKF-AH algorithm weights ω are used to calculate and obtain the accurate SOC estimate. NEW See the formula below.

[0121] SOC NEW (k)=ωSOC AH (k)+(1-ω)SOC EKF (k).

[0122] Furthermore, the corrected warning threshold δ is set to 10%.

[0123] Example

[0124] The research focused on lithium iron phosphate batteries. The experimental testing environment consisted of a PB5060LP28 STM32 DEMO, lithium iron phosphate batteries (pack), a BMS_Studio host computer, and an Yishengda battery testing system.

[0125] like Figure 1 As shown, a method for estimating and correcting the state of charge (SOC) of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm includes the following specific steps:

[0126] S1. Taking a specific power-on of the energy storage device as the starting point, the BMS reads the SOC and SOH values ​​from the last power-off, and records them as SOC. his SOH his .

[0127] S2. Calculate the time difference between the current power-on and the last power-off as the battery resting time ΔT, and determine whether the battery resting time ΔT has reached the preset battery resting time value T; in this embodiment, a lithium iron phosphate battery is used as an example, and T = 2h is taken.

[0128] If the battery resting time ΔT ≥ 2h, proceed to S3;

[0129] If the battery resting time ΔT < 2h, proceed to S4;

[0130] S3. Since the battery has been idle for a sufficiently long time, the SOC value is corrected and calculated using the open-circuit voltage OCV to obtain a new SOC. new And calculate SOC his With SOC new The difference ΔSOC=|SOC his -SOC new | Let the SOH correction warning threshold of the battery be δ. In this embodiment, the SOH correction warning threshold is δ = 10%.

[0131] If ΔSOC > 10%, then SOH is considered to be... his If the estimation is incorrect, proceed to step S9;

[0132] If ΔSOC≤10%, then proceed to step S5;

[0133] S3 uses the open-circuit voltage OCV to correct the SOC value, specifically including the following:

[0134] S3.1, Perform battery OCV-SOC test to obtain the relationship between battery open circuit voltage OCV and remaining capacity SOC under different temperatures and aging conditions;

[0135] S3.2, using MATLAB to fit the polynomial of the functional relationship between open-circuit voltage OCV and SOC.

[0136] OCV=f(SOC=a0+a1SOC 1 +a2SOC 2 +…+a n SOC n ,

[0137] Where a n The coefficients of the nth power term;

[0138] S3.3, Substitute the open-circuit voltage OCV read by the sensor after power-on into the polynomial to obtain SOC. new .

[0139] Due to factors such as aging, weather, and self-discharge, the open-circuit voltage (OCV) of a battery will continuously decrease during periods of inactivity. Therefore, SOC correction is necessary to reduce estimation errors caused by incorrect initial values. However, if the SOC correction value exceeds a threshold (in this embodiment, the threshold is set to 10%), it is considered that the battery aging problem has been verified, and an SOH correction is required to determine the fault.

[0140] S4. Due to the short battery resting time, the SOC value is corrected using the EKF algorithm. Noise correction is performed for 1 minute to obtain the corrected SOC. new Switch to S5.

[0141] S4 uses the EKF algorithm to correct the SOC calculation, specifically including the following:

[0142] S4.1, Based on the polarization characteristics of lithium batteries, establish the Thevenin equivalent circuit model, such as... Figure 2 As shown;

[0143] S4.2, the working equations of the Thevenin equivalent circuit model above are:

[0144]

[0145] In the above formula, OCV is the open-circuit voltage, R0 is the ohmic internal resistance of the battery, C is the polarization capacitor, R1 is the polarization resistor, and U1 represents the voltage across the polarization resistor R1 and the polarization capacitor C. s Where is the battery terminal voltage, and i is the current battery current, positive for charging and negative for discharging;

[0146] S4.3 Discretize the above equations and combine them with extended Kalman filtering to obtain the observation equations and measurement equations as follows.

[0147]

[0148] In the observation and measurement equations, t is the unit sampling interval, and Q... maxThe current maximum capacity of the battery is related to the state of equilibrium (SOH). Q represents the process noise error matrix, and R represents the measurement noise error matrix.

[0149] The algorithm recursively estimates as follows:

[0150] Prior estimation,

[0151]

[0152] Kalman gain update,

[0153]

[0154] Optimal estimate update,

[0155]

[0156] P is the prior estimate of the observed state at the current moment. k - Let K be the prior error covariance matrix at the current time. k Here is the Kalman gain matrix. P k + Let R be the optimal estimate and the updated error covariance matrix at the current time, Q be the process noise error matrix, and R be the error matrix. k Let U be the measurement noise error matrix at the current moment. s This is the value measured by the voltmeter. This is an estimate of the voltage at the opposite end at the current moment;

[0157] S4.4, HPPC test is performed, and the battery is charged and discharged using a continuous pulse sequence current to obtain the dynamic characteristics of the battery under different SOC values. MATLAB is used to identify the battery parameters offline to obtain R0, R1, and C.

[0158] S4.5, Substituting the previously estimated SOC(k-1) and U1(k-1), we obtain SOC. new =SOC(k).

[0159] For the sake of simplicity, this embodiment uses an offline parameter identification method. Therefore, S4.4 can be improved to various online parameter identification methods, and the results still fall within the protection scope of this patent.

[0160] The S4 step uses the EKF algorithm to correct the SOC. In order to improve the correction capability of the observation, the measurement noise error matrix R = 0.0003 is set in this step.

[0161] S5, Input the corrected SOC newUsing the EKF algorithm and the ampere-hour integration method as initial values, the SOC is estimated to obtain the SOC. EKF SOC AH By combining fuzzy control to establish a fuzzy rule base associated with the state of charge (SOC) and the state of health (SOH), the weights ω of the EKF-AH algorithm and the measurement noise error γ are dynamically adjusted to obtain an accurate estimate of SOC. NEW ;

[0162] like Figure 3 As shown, the OCV-SOC curve of lithium iron phosphate batteries exhibits a plateau period, and the OCV-SOC curves differ under different SOH conditions.

[0163] Based on actual test results, this embodiment selects [10%, 90%] as the SOC plateau period. Within this range, the OCV change is not significant, leading to a decrease in the accuracy of OCV correction and EKF estimation. At the same time, as the battery ages and the SOH decreases, the battery's OCV will change in the non-plateau period. Therefore, it is necessary to consider both whether the state of charge (SOC) is in the plateau period and the degree of battery aging characterized by the health status (lifetime) SOH.

[0164] S5 specifically includes the following:

[0165] In S5, fuzzy control dynamically adjusts the weights ω and measurement noise error γ of the EKF-AH algorithm, specifically including the following:

[0166] S5.1, Obtain SOC new SOH his Using the ampere-hour integration method and EKF to evaluate the SOC new SOC was estimated separately AH SOC EKF The discretization formula for the ampere-hour integral method is as follows:

[0167]

[0168] Q max =SOH his *Q 出厂 ,

[0169] In the above formula, t is the unit sampling interval, i is the current battery current (positive for charging, negative for discharging), and Q... max Q represents the battery's current maximum capacity. 出厂 Q represents the battery's maximum rated capacity at the time of manufacture. This capacity decreases as the battery is used. max The SOH level continues to decrease;

[0170] S5.2, in the fuzzy controller, set SOC new SOH hisAs input to the fuzzy controller, the EKF-AH algorithm weights ω and the measurement noise error γ are used as the system output;

[0171] Define three fuzzy subsets of SOC as Bigger(B), Middler(M), and Lower(L);

[0172] The three fuzzy subsets of SOH are defined as Old(O), NOW(M), and New(N);

[0173] Similarly, the fuzzy subsets of ω and γ are defined as Big(B), Mid(M), and Low(L), respectively.

[0174] Based on the membership function type and parameter settings, fuzzy inference rules are formulated through the design of the fuzzy controller, as shown in Table 1;

[0175] In this embodiment, the first input variable of fuzzy control is the state of charge (SOC). Here, the fuzzy subset is defined as [L, M, B], which correspond to the three intervals of SOC: [0%, 10%], (10%, 90%), and [90%, 100%], namely, the non-plateau period, the plateau period, and the non-plateau period.

[0176] The second variable is the health status (lifespan) SOH, which is defined as a fuzzy subset of [O, M, N], corresponding to the three intervals of SOH: [70%, 85%], (85%, 95%), and [95%, 100%), representing the old, medium, and new levels of battery aging.

[0177] Batteries with a SOH of less than 70% are usually recycled, so 70% is taken as the lower limit. The fuzzy domain and interval division can be adjusted according to the actual situation.

[0178] Similarly, the classification of ω and γ is also based on this principle. This section is only an example, and the specific classification needs to be tested for different battery specifications.

[0179] Table 1 Fuzzy Rule Base Table

[0180] ω and γ SOH=N SOH=M SOH=O SOC = B L M M SOC = M B B B SOC = L L M B

[0181] S5.3, through the design of a fuzzy controller, combined with the range of the State of Charge (SOC) and the magnitude of the State of Health (SOH), the fuzzy controller is used to achieve dynamic adaptive adjustment of the EKF-AH algorithm weights ω and the measurement noise error γ. Finally, the adjusted measurement noise error γ is passed to the measurement noise error matrix R in the Kalman filter gain coefficient, and the latest EKF-AH algorithm weights ω are used to calculate and obtain the accurate SOC estimate. NEW See the formula below.

[0182] SOC NEW (k)=ωSOCAH (k)+(1-ω)SOC EKF (k).

[0183] A larger ω indicates a greater weighting of the AH algorithm, suggesting higher reliability of the ampere-hour integration method. Table 1 shows that when the battery is in a plateau phase, ω and γ are chosen as Big(B) because the OCV-SOC curve exhibits poor nonlinearity in this case, making the algorithm more reliant on the ampere-hour integration method.

[0184] When battery aging causes SOH to decrease, the selection of ω and γ will increase, changing from L to M. This is because the OCV-SOC curve in this case is different from that of the battery at the factory, which makes the OCV estimation part of the Kalman filter inaccurate, and the algorithm trusts the ampere-hour integration method more.

[0185] In this embodiment, since the battery is in good health and SOH = N, when it is in the plateau period (SOC = M), ω(B) = 0.7 and γ(B) = 0.03; when it is in the non-plateau period (SOC = L or B), ω(L) = 0.45 and γ(L) = 0.0003.

[0186] S6, calculate the estimation difference e(k) = |SOC| between the EKF algorithm and the ampere-hour integration method. EKF -SOC AH Establish TS fuzzy control rules and set different warning thresholds α based on whether SOC is in a plateau period.

[0187] When 10% < SOC NEW If the SOC is less than 90%, meaning it is in a plateau phase, then the warning threshold α is set to 2%.

[0188] When SOC NEW ≤10%, or SOC NEW If the SOC is ≥90%, meaning it is in a non-plateau period, then the warning threshold α is set to 5%.

[0189] S7 is obtained by summing the differences from the first n estimations. If the sum of the estimated differences e(k) is greater than the corresponding warning threshold α, proceed to step S9; otherwise, proceed to step S8.

[0190] S8, BMS updates the current SOC of energy storage devices, providing an accurate SOC estimate. NEW Assign to SOC new Proceed to step S5;

[0191] S9, BMS identifies SOH his An estimate was incorrect; a SOH correction warning has been issued.

[0192] Before algorithm fusion, the EKF algorithm and the ampere-hour integration method must have estimation errors, and the error gap between the two algorithms is different in different SOC ranges.

[0193] Since the EKF algorithm will rely more on the ampere-hour integration method during the plateau period, a larger measurement noise error γ is selected. Theoretically, the algorithm error should be smaller, so a smaller warning threshold α = 2% is selected.

[0194] Similarly, the slope of the SOC-OCV curve is larger during the non-plateau period. The EKF algorithm requires a smaller measurement noise error γ. Theoretically, the algorithm error should be larger, so the warning threshold α = 5% is selected.

[0195] Since both algorithms rely on the ampere-hour integration method to varying degrees, and the ampere-hour integration method is affected by the state of health (SOH), if the error of either algorithm exceeds the threshold, it indicates that the currently displayed SOH result for battery health is inaccurate and needs to be re-estimated and verified.

[0196] like Figure 4 , 5 As shown, this embodiment performs a single charging test, estimating the State of Charge (SOC) using both the single EKF algorithm and the EKF-AH algorithm, the latter employing dynamic parameter adjustment. The figure includes the estimation error, the SOC estimation result, and the terminal voltage estimation result. The more accurate the estimation result, the more the graphs of the actual and estimated values ​​overlap. From... Figure 4 As can be seen, the estimation error of the single EKF algorithm is relatively large, so the actual value and the estimated value do not overlap; from Figure 5 As can be seen, the EKF-AH algorithm has good robustness, high curve overlap, and is more accurate.

[0197] The results show that the estimation error of the single EKF algorithm stabilizes at 3%, while the estimation error based on the fuzzy EKF-AH algorithm stabilizes at around 0.2%.

[0198] In summary, the EKF-AH algorithm, which combines extended Kalman filtering and ampere-hour integration, is used to estimate the State of Charge (SOC) of lithium iron phosphate (LFP) batteries. This eliminates the initial value error and cumulative process error of the ampere-hour integration method and solves the problem of inaccuracy of EKF during the plateau period. A fuzzy rule base is established that is associated with the current SOC and the State of Health (SOH) to dynamically adjust the weights of the EKF-AH algorithm and the measurement noise error γ, thereby improving the estimation accuracy of LFP batteries throughout their entire life cycle. Different thresholds are set according to whether the SOC is in a plateau period. The cumulative estimation error of the two algorithms is compared with the threshold to determine whether the SOH estimation is incorrect, thus assisting in the correction of the SOH.

[0199] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A method for estimating and correcting the state of charge (SOC) of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm, characterized in that, Includes the following steps, S1, starting with a specific power-on of the energy storage device, the BMS reads the SOC and SOH values ​​from the last power-off, and records them as SOC. his SOH his ; S2, calculate the time difference between the current power-on and the last power-off as the battery rest time ΔT, and determine whether the battery rest time ΔT has reached the preset value T of the battery rest time; If the battery resting time ΔT≥T, proceed to step S3; If the battery resting time ΔT < T, proceed to step S4; S3, use the open-circuit voltage OCV to correct the SOC value and obtain the new SOC. new And calculate SOC his With SOC new The difference ΔSOC=|SOC his -SOC new Let the SOH correction warning threshold for the battery be δ; If ΔSOC>δ, determine SOH. his If the estimation is incorrect, proceed to step S9; If ΔSOC≤δ, then proceed to step S5; S4, use the EKF algorithm to correct the SOC. The correction process lasts for 1 minute, yielding the corrected SOC. new Proceed to step S5; S5, input SOC new Using the EKF algorithm and the ampere-hour integration method as initial values, the SOC of the next state is estimated, and the SOC is obtained. EKF SOC AH By combining fuzzy control to establish a fuzzy rule base associated with the state of charge (SOC) and the state of health (SOH), the weights ω of the EKF-AH algorithm and the measurement noise error γ are dynamically adjusted to obtain an accurate estimate of SOC. NEW ; S6, calculate the estimation difference e(k) = |SOC| between the EKF algorithm and the ampere-hour integration method. EKF -SOC AH Establish TS fuzzy control rules and set different warning thresholds α based on whether SOC is in a plateau period. When 10% < SOC NEW If the SOC is less than 90%, meaning it is in a plateau phase, then the warning threshold α is set to 2%. When SOC NEW ≤10%, or SOC NEW If the SOC is ≥90%, meaning it is in a non-plateau period, then the warning threshold α is set to 5%. S7 is obtained by summing the differences from the first n estimations. If the sum of the estimated differences e(k) is greater than the corresponding warning threshold α, proceed to step S9; otherwise, proceed to step S8. S8, BMS updates the current SOC of energy storage devices, providing an accurate SOC estimate. NEW Assign to SOC new Proceed to step S5; S9, BMS determines SOH his An estimate was incorrect; a SOH correction warning has been issued.

2. The method for estimating and correcting the SOC of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm according to claim 1, characterized in that, The S3 step uses the open-circuit voltage OCV to correct the SOC value, specifically including the following: S3.1, Perform battery OCV-SOC test to obtain the relationship between battery open circuit voltage OCV and remaining capacity SOC under different temperatures and aging conditions; S3.2, using MATLAB to fit the polynomial of the functional relationship between open-circuit voltage OCV and SOC. OCV=f(SOC)=a0+a1SOC 1 +a2SOC 2 +…+a n SOCIETY n , Where a n The coefficients of the nth power term; S3.3, Substitute the open-circuit voltage OCV read by the sensor after power-on into the polynomial to obtain SOC. new .

3. The method for estimating and correcting the SOC of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm according to claim 1, characterized in that, The S4 step uses the EKF algorithm to correct the SOC calculation, specifically including the following: S4.1, Based on the polarization characteristics of lithium batteries, establish the Thevenin equivalent circuit model; S4.2, the working equations of the Thevenin equivalent circuit model above are: In the above formula, OCV is the open-circuit voltage, R0 is the ohmic internal resistance of the battery, C is the polarization capacitor, R1 is the polarization resistor, and U1 represents the voltage across the polarization resistor R1 and the polarization capacitor C. s Where is the battery terminal voltage, and i is the current battery current, positive for charging and negative for discharging; S4.3 Discretize the above equations and combine them with extended Kalman filtering to obtain the observation equations and measurement equations as follows. In the observation and measurement equations, t is the unit sampling interval, and Q... max The current maximum capacity of the battery is related to the state of equilibrium (SOH). Q represents the process noise error matrix, and R represents the measurement noise error matrix. The algorithm recursively estimates as follows: Prior estimation, Kalman gain update, Optimal estimate update, P is the prior estimate of the observed state at the current moment. k - Let K be the prior error covariance matrix at the current time. k Here is the Kalman gain matrix. P k + Let R be the optimal estimate and the updated error covariance matrix at the current time, Q be the process noise error matrix, and R be the error matrix. k Let U be the measurement noise error matrix at the current moment. s This is the value measured by the voltmeter. This is an estimate of the voltage at the opposite end at the current moment; S4.4, HPPC test is performed, and the battery is charged and discharged using a continuous pulse sequence current to obtain the dynamic characteristics of the battery under different SOC values. MATLAB is used to identify the battery parameters offline to obtain R0, R1, and C. S4.5, Substituting the previously estimated SOC(k-1) and U1(k-1), we obtain SOC. new =SOC(k).

4. The method for estimating and correcting the SOC of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm according to claim 3, characterized in that, The S4 uses the EKF algorithm to correct the SOC, and its measurement noise error matrix R = 0.0003.

5. The method for estimating and correcting the SOC of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm according to claim 1, characterized in that, The fuzzy control in S5 dynamically adjusts the weights ω and measurement noise error γ of the EKF-AH algorithm, specifically including the following: S5.1, Obtain SOC new SOH his Using the ampere-hour integration method and EKF to evaluate the SOC new SOC was estimated separately AH SOC EKF The discretization formula for the ampere-hour integral method is as follows: Q max =SOH his *Q 出厂 , In the above formula, t is the unit sampling interval, i is the current battery current (positive for charging, negative for discharging), and Q... max Q represents the battery's current maximum capacity. 出厂 Q represents the battery's maximum rated capacity at the time of manufacture. This capacity decreases as the battery is used. max The SOH level continues to decrease; S5.2, in the fuzzy controller, set SOC new SOH his As input to the fuzzy controller, the EKF-AH algorithm weights ω and the measurement noise error γ are used as the system output; Define three fuzzy subsets of SOC as Bigger(B), Middler(M), and Lower(L); The three fuzzy subsets of SOH are defined as Old(O), NOW(M), and New(N); Similarly, the fuzzy subsets of ω and γ are defined as Big(B), Mid(M), and Low(L), respectively. Based on the membership function type and parameter settings, fuzzy inference rules are formulated through the design of a fuzzy controller; S5.3, through the design of a fuzzy controller, combined with the range of the State of Charge (SOC) and the magnitude of the State of Health (SOH), the fuzzy controller is used to achieve dynamic adaptive adjustment of the EKF-AH algorithm weights ω and the measurement noise error γ. Finally, the adjusted measurement noise error γ is passed to the measurement noise error matrix R in the Kalman filter gain coefficient, and the latest EKF-AH algorithm weights ω are used to calculate and obtain the accurate SOC estimate. NEW See the formula below. SOC NEW (k)=ωSOC AH (k)+(1-ω)SOC EKF (k)。 6. The method for estimating and correcting the SOC of lithium iron phosphate batteries based on the fuzzy EKF-AH algorithm according to claim 1, characterized in that, The corrected warning threshold δ is set to 10%.

Images