Series battery pack energy state estimation method

Abstract

The invention discloses a series battery pack energy state estimation method, which comprises the following specific steps of: firstly, measuring a battery pack end voltage Ub, obtaining a battery pack charge state initial value SOC0 through a table look-up method, and calculating to obtain a battery pack energy state initial value SOE0; a recursive least square method is used to identify battery model parameters, and standardized processing is carried out on the identified voltage Ui and internal resistance R0i; in order to solve the problem of battery performance inconsistency, an elbow method is adopted to analyze the battery characteristic difference, the Euclidean distance D is converted into a similarity value S, the sum of squares SSE is calculated, a relation curve between the sum of squares SSE and an inconsistency threshold T is drawn, and the optimal threshold T is determined through the elbow position; if the characteristics of the batteries are similar, K-means clustering is used; if the characteristics of the batteries are not similar, fuzzy C-Means clustering is used; and calculating the maximum available energy En of the battery pack based on a clustering result, substituting the En and the SOE0 into an adaptive extended Kalman filter to carry out energy state estimation, and finally obtaining an energy state value SOEr of the battery pack.

Description

Technical Field

[0001] The present invention belongs to the technical field of design and control of large-capacity battery energy storage systems in smart grids, and relates to a method for estimating the energy state of a series-connected battery pack. Background Art

[0002] With the rapid development of battery technology, series-connected battery packs are widely used in electric vehicles and renewable energy storage because they can effectively increase voltage and energy output. However, when multiple battery cells are connected in series, the inconsistency problem between batteries becomes more and more prominent, including manufacturing differences, aging degree and different working environments. These inconsistencies lead to uneven performance of each cell during the charging and discharging process, which directly affects the estimation of the state of energy (SOE) of the entire battery pack. Accurately estimating the SOE of series-connected battery packs is crucial for optimizing energy management, improving safety and extending service life. Traditional SOE estimation methods often have difficulty in effectively solving the inconsistency problem between cells. Therefore, it is particularly important to develop new estimation methods to achieve accurate prediction of the energy state in series-connected battery packs. This will help users better plan and maintain battery packs, thereby improving their overall performance and reliability.

[0003] The patent (CN115859567A) discloses a dual-time scale fuzzy extended Kalman filter (EKF) multi-state joint estimation method. n Estimation problem, this method uses the forgetting factor recursive least squares (RLS) algorithm to perform online parameter identification, effectively deal with the data saturation problem, and improve the accuracy of the equivalent model. The multi-state joint estimation method of lithium batteries based on dual-time scale fuzzy EKF first accurately estimates the model parameters. Subsequently, the functional relationship between the battery open circuit voltage and SOE is dynamically adjusted through the fuzzy controller to adapt to the changes in the battery charge and discharge current rate, thereby improving the estimation accuracy of the lithium battery SOE. This method adopts a small time scale sampling strategy to accurately estimate the battery SOE. After completing multiple SOE estimates, a large time scale sampling method is used to estimate the maximum available energy E of the battery. n , which aims to reduce the volatility of the maximum available energy estimation results caused by SOE mutations and achieve accurate estimation of the maximum available energy of the battery. However, the invention has not fully considered the inconsistency problem between battery cells, which may affect SOE and maximum available energy E n To solve this problem, the present invention proposes to set the inconsistency threshold by measuring the voltage and internal resistance of the battery cells, thereby judging the performance difference between the battery cells, and selecting a suitable clustering method, combined with the AEKF algorithm, to achieve accurate estimation of the energy state of the battery pack. Summary of the invention

[0004] The problem solved by the present invention is to provide a method for estimating the energy state of a series battery pack. First, by selecting voltage and internal resistance as key parameters, calculating the Euclidean distance between batteries and calculating the similarity value S, and setting the inconsistency threshold T to quantify the difference in battery characteristics within the battery pack, the problem of how to evaluate the inconsistency of battery characteristics is solved. Second, according to the similarity of battery characteristics, the K-Means or fuzzy C-Means clustering algorithm is flexibly selected, and representative batteries are selected by the clustering method, combined with the maximum available energy E n The AEKF algorithm is used to dynamically correct the battery SOE, which solves the problem of large errors in the estimation of the energy state of the battery pack and improves the overall performance of the battery management system. To achieve the above purpose, the present invention provides the following technical solutions: a method for estimating the energy state of a series battery pack, the specific steps are as follows:

[0005] S1. Measure and obtain battery pack and battery cell parameter data (1), including battery pack terminal voltage U b , battery pack current I, battery cell open circuit voltage U 0i , each battery cell terminal voltage U i 、Battery monomer internal resistance R 0i , where i is the serial number of the i-th battery cell;

[0006] S2, the battery terminal voltage U b The initial value SOC of the battery pack state of charge is obtained by table lookup method 0 , combined with the battery pack rated energy E rated Calculate the initial value of the battery pack energy state SOE 0 (2);

[0007] S3, the measured terminal voltage U of each battery cell i 、Battery monomer internal resistance R 0i Perform parameter standardization (3), calculate the Euclidean distance D (4), and convert D into a similarity value S; use the standardized battery parameter data and the elbow method (5) to obtain the optimal threshold T for battery inconsistency, then compare S with T. If S is less than or equal to T, it indicates that the battery pack characteristics are similar. Use the K-Means clustering method (6) to perform hard clustering, determine the representative battery, and calculate the maximum available energy E of the battery pack. n (8); If S is greater than T, it indicates that the battery pack characteristics are not similar. The fuzzy C-Means clustering method (7) is used to determine the representative battery and calculate the maximum available energy E of the battery pack. n(8); The elbow method is designed as follows: first, set the range of the battery U and R inconsistency threshold T, then calculate the sum of squared errors (SSE) between each data point (battery) and its corresponding cluster center, and then draw the relationship curve between SSE and threshold T, find the "elbow" position of the curve, and determine the optimal threshold T;

[0008] S4, with battery pack current I, battery cell open circuit voltage U 0i , each battery cell terminal voltage U i 、Battery monomer internal resistance R 0i As input, the recursive least square method is used to identify the battery cell parameters online (9), and the equivalent model of the series battery pack is established in combination with the series circuit characteristics (10);

[0009] S5. Establish the space state equation through the battery pack equivalent model (10), and convert the initial value of the battery pack energy state SOE 0 and the maximum available energy of the battery pack E n Substitute the adaptive extended Kalman filter into the energy state of the series battery pack using the AEKF algorithm (11), and finally obtain its energy state value SOE r .

[0010] The battery similarity determination method process in step S3 is specifically as follows:

[0011] (1) Calculate the Euclidean distance D(4) and convert D into a similarity value S. The Euclidean distance formula is: Among them, the subscripts 1 and 2 represent different battery cells, D 1,2 is the Euclidean distance between battery 1 and battery 2, U' 1 and U' 2 , R' 01 and R' 02 They are the standardized terminal voltage and internal resistance of battery 1 and 2 respectively, and finally the formula This finds the similarity S;

[0012] (2) Set an inconsistency threshold T to determine whether the battery characteristics are different enough. Use the elbow method (5) to calculate the error sum of squares SSE under different cluster numbers K to evaluate the clustering effect. The error sum of squares formula is: Where M is a specific group of batteries, U' i and R' 0i are the standardized terminal voltage and internal resistance of the ith battery, and are the mean values ​​of the voltage and internal resistance of the batteries in the group (i.e., the center of the group), and then the SSE curves under different thresholds T are plotted to find the "elbow" point as the optimal threshold T, i.e., where the SSE decrease rate is significantly reduced;

[0013] (3) Make the following judgment based on the calculated similarity value S and the set threshold T: if the similarity between batteries is less than or equal to the set threshold T, that is, S≤T, then it is considered that the characteristics of these batteries are similar and can be clustered into the same group; if the similarity between batteries is greater than the set threshold T, that is, S>T, then it is considered that the characteristics of these batteries are not similar and further fuzzy clustering processing is required.

[0014] In step S3, different clustering methods are used according to battery inconsistency. The specific steps are as follows:

[0015] (1) When the battery characteristics are similar, use the K-Means clustering method (6), select the number of clusters K, randomly select K initial cluster centers, assign each battery to the nearest cluster center, then update the cluster center, and repeat the iteration until convergence. The clustering formula is: Among them C j and S j are the center of the jth cluster and the point set belonging to the jth cluster, respectively. Finally, the battery closest to the cluster center is selected from each cluster as the representative battery and the maximum available energy E of the battery pack is calculated. n (8);

[0016] (2) When the battery characteristics are not similar, the fuzzy C-means clustering method (7) is used. The initial number of clusters C is selected based on data complexity or prior knowledge. The membership degree of each battery is calculated, that is, the probability that the battery belongs to each cluster. The calculation formula is: where u ij is the membership degree of the ith battery to the jth cluster, d ij is the distance between the ith battery and the jth battery cluster center, m is the fuzzy index, and then the cluster center is updated. The update formula is: Repeat the iterative update until the cluster center converges, divide the batteries into multiple clusters, and output the membership degree of each battery. Finally, find the battery with the highest membership degree in each cluster as the representative battery and calculate the maximum available energy E of the battery pack. n (8).

[0017] The adaptive extended Kalman filter (AEKF) algorithm (11) described in step S5 has the following specific steps: ① initializing state variables, mean and covariance; ② time updating of state estimation; ③ time updating of covariance error; ④ calculating Kalman gain; ⑤ measurement updating of state estimation; ⑥ measurement updating of covariance error; ⑦ online parameter adjustment; ⑧ loop iteration; through the above steps, the initial value SOE of the battery energy state is set to 0 and the maximum available energy E of the battery pack n Substitute it into the AEKF algorithm (11) for loop iteration to obtain the battery pack energy state value SOE in real time r.

[0018] Compared with the disclosed patent (CN115859567A), the present invention has the following beneficial effects: first, by selecting voltage and internal resistance as key parameters, calculating the Euclidean distance D between batteries and converting it into a similarity value S, setting an inconsistency threshold T to quantify the differences in battery characteristics within a battery pack, thereby effectively solving the problem of evaluating the inconsistency of battery characteristics; second, according to the similarity of battery characteristics, flexibly selecting K-Means or fuzzy C-Means clustering algorithms, determining representative batteries through clustering methods, and combining the initial value of the battery energy state SOE 0 and the maximum available energy E n , using the AEKF algorithm to dynamically correct the battery pack energy state value SOE r , which solves the problem of large error in battery pack energy state estimation and significantly improves the overall performance of the battery management system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 A structural diagram of an energy state estimation method for a series battery pack;

[0020] Figure 2 It is the equivalent circuit model diagram of the battery cell;

[0021] Figure 3 It is the equivalent circuit model diagram of series battery pack;

[0022] Figure 4 This is a structural diagram of a series lithium-ion battery pack. DETAILED DESCRIPTION

[0023] The present invention is further described in detail below in conjunction with specific examples, which are intended to explain the present invention rather than to limit it.

[0024] Example: Figure 1 As shown, a method for estimating the energy state of a series battery pack is provided. The series battery pack is formed by connecting a plurality of battery cells in series. The system is formed by connecting n battery cells in series, where n is a natural number greater than or equal to 1. Figure 4 As shown, the specific steps are as follows:

[0025] The series battery pack is composed of n=3 battery cells connected in series. Figure 3 As shown; the rated voltage of the battery cell is 3.7V, the discharge cut-off voltage is 3V, and the rated capacity is 860mAh.

[0026] S1. Measure and obtain battery pack and battery cell parameter data (1), including battery pack terminal voltage U b , battery pack current I, battery cell open circuit voltage U 0i , each battery cell terminal voltage U i、Battery monomer internal resistance R 0i , where i is the serial number of the i-th battery cell;

[0027] The measured battery terminal voltage U b =10.72V, battery pack current I = 1.0A, battery cell open circuit voltage U 0i =[3.7,3.65,3.6]V, each battery cell terminal voltage U i =[3.6,3.55,3.57]V, battery cell internal resistance R 0i =[0.05,0.04,0.06]Ω, i=1,2,3.

[0028] S2, the battery terminal voltage U b The initial value SOC of the battery pack state of charge is obtained by table lookup method 0 , combined with the battery pack rated energy E rated Calculate the initial value of the battery pack energy state SOE 0 (2), the specific formula is: E rated =U rated ×C rated , SOE 0 =SOC 0 ×E rated , where U rated and C rated They are the rated voltage and rated capacity of the battery pack respectively.

[0029] S3, the measured terminal voltage U of each battery cell i 、Battery monomer internal resistance R 0i Perform parameter standardization (3), calculate the Euclidean distance D (4), and convert D into a similarity value S; use the standardized battery parameter data and use the elbow method (5) to obtain the optimal threshold T for battery inconsistency, then compare S with T. If S is less than or equal to T, it indicates that the battery pack characteristics are similar. Use the K-Means clustering method (6) to perform hard clustering, determine the representative battery, and calculate the maximum available energy E of the battery pack. n (8); If S is greater than T, it indicates that the battery pack characteristics are not similar. The fuzzy C-Means clustering method (7) is used to determine the representative battery and calculate the maximum available energy E of the battery pack. n (8); The elbow method is designed as follows: first, set the range of the battery U and R inconsistency threshold T, then calculate the sum of squared errors (SSE) between each data point (battery) and its corresponding cluster center, and then draw the relationship curve between SSE and threshold T, find the "elbow" position of the curve, and determine the optimal threshold T; the specific implementation steps are as follows:

[0030] (1) Calculate the terminal voltage U' of each battery cell after parameter standardizationi =[0.974,0.949,0.935], battery cell internal resistance R' 0i =[0.833,0.667,1.000];

[0031] (2) Calculated by the Euclidean distance formula:

[0032]

[0033] (3) Convert D into similarity value S:

[0034]

[0035] (4) Using the elbow method to obtain the optimal battery inconsistency threshold T = 0.8

[0036] (5) Compare S with T:

[0037]

[0038] Therefore, we can conclude that the characteristics of battery 2 and battery 3 are similar, the characteristics of battery 1 and battery 2 are not similar, and the characteristics of battery 1 and battery 3 are not similar. Since battery 2 and battery 3 are judged to be similar and there is no fuzzy overlap, they can be clearly classified into the same cluster and clustered using the K-means clustering method; battery 1 has overlapping features with batteries 2 and 3, but they are not completely similar, so they are clustered using the fuzzy C-means clustering method;

[0039] (6) Using the K-means clustering method, the battery closest to the cluster center is selected as the representative battery; using the fuzzy C-means clustering method, the battery with the highest membership in each cluster is selected as the representative battery according to the membership degree; finally, the minimum value method is used to calculate the maximum available energy E of the battery pack. n .

[0040] S4, with battery pack current I, battery cell open circuit voltage U 0i , each battery cell terminal voltage U i 、Battery monomer internal resistance R 0i As input, the recursive least square method is used to identify the battery cell parameters online (9), and the equivalent model of the series battery pack is established in combination with the series circuit characteristics (10); the series battery pack is composed of n=3 battery cells connected in series, the rated voltage of the battery cell is 3.7V, the discharge cut-off voltage is 3V, and the rated capacity is 860mAh.

[0041] The equivalent circuit model of the battery cell is a model containing a second-order RC circuit, which consists of a controlled voltage source U 0 (SOC), internal resistance R 0 (t), electrochemical polarization resistance Rs (t), electrochemical polarization capacitance C s (t) and concentration polarization resistance R l (t), concentration polarization capacitance C l (t) composition, the terminal voltage is

[0042] U t (t) = U 0 [SOC(t)]-I(t)[R 0 (t)+R s (t) / R s (t)jωC s (t)+R l (t) / R l (t)jωC l (t)]

[0043] The relationship between the above parameters and SOC is:

[0044]

[0045] In the formula, the model coefficient a obtained by fitting the battery experimental data is 0 ~a 5 The values ​​are -0.915, -40.867, 3.632, 0.537, -0.499, 0.522, respectively. 0 ~b 5 The values ​​are -0.1463, -30.27, 0.1037, 0.0584, 0.1747, 0.1288, respectively. 0 ~c 2 The values ​​are 0.1063, -62.49, 0.0437, respectively. 0 ~d 2 The values ​​are -200, -138, 300, e 0 ~e 2 The values ​​are 0.0712, -61.4, 0.0288, respectively. 0 ~f 2 The values ​​are -3083, -180, and 5088 respectively.

[0046] According to the above mathematical model, the equivalent circuit model of the battery cell is established with SOC c , U s and U l is the state vector, and the spatial state equation is obtained as:

[0047]

[0048] [U t,k ]=U o,k -R 0,k Ik -U s,k -U l,k +ν k

[0049] In the formula, ω k is the process noise, ν k To measure noise.

[0050] S5. Establish the space state equation through the battery pack equivalent model (10), and convert the initial value of the battery pack energy state SOE 0 and the maximum available energy of the battery pack E n Substitute the adaptive extended Kalman filter into the energy state of the series battery pack using the AEKF algorithm (11), and finally obtain its energy state value SOE r ;

[0051] The specific steps of the adaptive extended Kalman filter (AEKF) algorithm (11) are as follows: ① Initialize state variables, mean and covariance; ② Time update of state estimation; ③ Time update of covariance error; ④ Calculate Kalman gain; ⑤ Measurement update of state estimation; ⑥ Measurement update of covariance error; ⑦ Online parameter adjustment; ⑧ Loop iteration; Through the above steps, the initial value of battery energy state SOE is 0 and the maximum available energy E of the battery pack n Substitute it into the AEKF algorithm (11) for loop iteration to obtain the battery pack energy state value SOE in real time r .

Claims

1. A method for estimating the energy state of a series battery pack, wherein the series battery pack is composed of n battery cells connected in series, wherein n is a natural number greater than 1, and the method comprises the following steps: S1. Measure and obtain the battery pack and battery cell parameter data, including the battery pack terminal voltage U b , battery pack current I, battery cell open circuit voltage U 0i , each battery cell terminal voltage U i 、Battery monomer internal resistance R 0i , where i is the serial number of the i-th battery cell; S2, the battery terminal voltage U b The initial value of the battery pack state of charge SOC0 is obtained by table lookup method, and then combined with the rated energy E of the battery pack rated The initial value of the battery pack energy state SOE0 is calculated; S3, the measured terminal voltage U of each battery cell i 、Battery monomer internal resistance R 0i Perform parameter standardization, calculate the Euclidean distance D, and convert D into a similarity value S; use the standardized battery parameter data, use the elbow method to obtain the optimal threshold T for battery inconsistency, and then compare S with T. If S is less than or equal to T, it indicates that the battery pack characteristics are similar. Use the K-Means clustering method for hard clustering to determine the representative battery and calculate the maximum available energy En of the battery pack; if S is greater than T, it indicates that the battery pack characteristics are not similar. Use the fuzzy C-Means clustering method to determine the representative battery and calculate the maximum available energy E of the battery pack n ; The elbow method is designed as follows: first set the range of the battery U and R inconsistency threshold T, then calculate the sum of squared errors (SSE) between each data point (battery) and its corresponding cluster center, then draw the relationship curve between SSE and threshold T, find the "elbow" position of the curve, and determine the optimal threshold T; S4, with battery pack current I, battery cell open circuit voltage U 0i , each battery cell terminal voltage U i 、Battery monomer internal resistance R 0i As input, the recursive least square method is used to identify the battery cell parameters online, and the equivalent model of the series battery pack is established in combination with the series circuit characteristics. S5, establish the space state equation through the battery pack equivalent model, and transform the initial value of the battery pack energy state SOE0 and the maximum available energy E of the battery pack n Substitute it into the adaptive extended Kalman filter AEKF, use AEKF to estimate the energy state of the series battery pack, and finally get its energy state value SOE r .

2. The method for estimating the energy state of a series battery pack according to claim 1, characterized in that: The battery similarity determination method process is as follows: (1) Calculate the Euclidean distance D and convert D into a similarity value S. The Euclidean distance formula is: Among them, the subscripts 1 and 2 represent different battery cells, D 1,2 is the Euclidean distance between battery 1 and battery 2, U1' and U'2, R' 01 and R' 02 They are the standardized terminal voltage and internal resistance of battery 1 and 2 respectively, and finally the formula This finds the similarity S; (2) Set an inconsistency threshold T to determine whether the battery characteristics are different enough. Use the elbow method to calculate the sum of squared errors (SSE) under different cluster numbers K to evaluate the clustering effect. The formula for the sum of squared errors is: Where M is a specific group of batteries, U i 'and R' 0i are the standardized terminal voltage and internal resistance of the ith battery, and are the terminal voltage and the mean value of the internal resistance of the battery in the group (i.e., the group center), and then the SSE curves under different thresholds T are plotted to find the "elbow" point as the optimal threshold T, i.e., where the SSE decrease rate is significantly reduced; (3) Make the following judgment based on the calculated similarity value S and the set threshold T: if the similarity between batteries is less than or equal to the set threshold T, that is, S≤T, then it is considered that the characteristics of these batteries are similar and can be clustered into the same group; if the similarity between batteries is greater than the set threshold T, that is, S>T, then it is considered that the characteristics of these batteries are not similar and further fuzzy clustering processing is required.

3. The method for estimating the energy state of a series battery pack according to claim 1, characterized in that: Different clustering methods are used when the battery characteristics are similar or dissimilar. The specific steps are as follows: (1) When the battery characteristics are similar, the K-Means clustering method is used. The number of clusters K is selected, K initial cluster centers are randomly selected, and each battery is assigned to the nearest cluster center. Then the cluster center is updated and iterated repeatedly until convergence. The clustering formula is: Among them C j and S j are the center of the jth cluster and the set of points belonging to the jth cluster, respectively. Finally, the battery closest to the cluster center is selected from each cluster as the representative battery; (2) When the battery characteristics are not similar, the fuzzy C-Means clustering method is used. The initial number of clusters C is selected based on data complexity or prior knowledge. The membership degree of each battery is calculated, that is, the probability that the battery belongs to each cluster. The calculation formula is: where u ij is the membership degree of the ith battery to the jth cluster, d ij is the distance between the ith battery and the jth battery cluster center, m is the fuzzy index, and then the cluster center is updated. The update formula is: Repeat the iterative update until the cluster centers converge, divide the batteries into multiple clusters, and output the membership of each battery, and finally find the battery with the highest membership in each cluster as the representative battery.

4. The method for estimating the energy state of a series battery pack according to claim 1, characterized in that: Combining the maximum available energy E through adaptive extended Kalman filter AEKF n And the initial value of the battery pack energy state SOE0 to achieve accurate estimation of the energy state of the series battery pack. The specific steps are as follows: (1) Based on the battery pack current I and the battery cell open circuit voltage U 0i , each battery cell terminal voltage U i 、Battery monomer internal resistance R 0i As input, the recursive least square method is used to identify the battery cell parameters online, and the equivalent model of the series battery pack is established in combination with the series circuit characteristics. (2) The space state equation is established through the battery pack equivalent model, and the initial value of the battery pack energy state SOE0 and the maximum available energy of the battery pack E n Substitute the adaptive extended Kalman filter, initialize the state variables and covariance matrix, then update the state estimate and covariance error time, then calculate the Kalman gain, and combine the battery pack voltage measurement to update the state estimate and covariance error; use the AEKF algorithm to estimate the energy state of the series battery pack, and finally get its energy state value SOE r .

Images