A method for estimating the actual capacity of lithium-ion batteries under non-perfect constant voltage conditions.

Abstract

The application discloses a kind of lithium ion battery actual capacity estimation methods applied to non-complete constant voltage working condition, comprising: offline determination battery model structure;Identify the battery model parameters corresponding to different data lengths;Further determine the effective data length interval, while determining the characteristic variable of the battery capacity attenuation and normalizing;Build the battery actual capacity estimation model corresponding to different constant voltage charging data length and store;Real-time record and store battery constant voltage charging time-current sequence;After the end of charging process, the characteristic variable is identified and normalized conversion is carried out;Query the battery actual capacity estimation model corresponding to actual constant voltage charging data;The actual capacity value of battery is calculated.The estimation method provided by the application has higher robustness compared with the battery actual capacity estimation method based on dynamic discharge data and constant current charging data.

Description

Technical Field

[0001] This invention belongs to the field of battery technology, and in particular relates to a method for estimating the actual capacity of lithium-ion batteries under non-perfect constant voltage conditions. Background Technology

[0002] Efficient and accurate estimation of the State of Health (SoH) of lithium-ion batteries is crucial for the safe and reliable operation of battery systems. Existing battery SoH estimation methods mainly fall into two categories: model-based and data-driven estimation methods. Model-based estimation methods primarily estimate the actual battery capacity based on electrochemical models, equivalent circuit models, or empirical models, combined with relevant state observers or filters. The estimation performance of this type of method is closely related to the model structure used; higher-order battery models can more accurately characterize battery chemical reactions and improve the accuracy of battery state estimation; however, they also increase the computational burden.

[0003] Based on the battery's operating conditions, data-driven estimation methods can be further divided into estimation methods based on charging and discharging conditions. Compared to discharging conditions, battery charging data is more stable and predictable. Therefore, battery actual capacity estimation methods based on charging operation data have received widespread attention. Among these, constant voltage charging conditions, due to their insensitivity to the initial charging state and the presence of more information reflecting battery capacity degradation, have been increasingly used in battery actual capacity estimation in recent years. For these methods, the accuracy of battery actual capacity estimation is closely related to the characteristic variables used, and the hyperparameters of the battery capacity estimation model and the completeness of the constant voltage charging data also have a significant impact on the estimation results. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method for estimating the actual capacity of lithium-ion batteries under non-complete constant-voltage conditions. This method accurately estimates the actual capacity of the battery online based on an equivalent circuit model of the battery under constant-voltage conditions after the battery has completed its constant-voltage charging process, thereby achieving flexible and accurate estimation of the battery's health status under non-complete charging conditions.

[0005] The present invention is achieved through the following technical solution.

[0006] This invention provides a method for estimating the actual capacity of lithium-ion batteries under non-perfect constant voltage conditions, the method comprising the following steps:

[0007] S1. Determine the battery model structure based on the test data;

[0008] S2, based on the battery's constant voltage charging current curve, identify the battery model parameters corresponding to different data lengths;

[0009] S3, perform correlation analysis between battery model parameters and battery capacity for different data lengths throughout the entire life cycle, further determine the effective data length range, and at the same time determine and normalize the characteristic variables that characterize battery capacity decay.

[0010] S4, combining optimization algorithms and machine learning algorithms, constructs and stores a battery actual capacity estimation model corresponding to different constant voltage charging data lengths;

[0011] S5. After the battery enters the constant voltage charging mode, record and store the battery constant voltage charging time-current sequence.

[0012] S6. After the charging process is completed, the characteristic variables determined in step S3 are identified based on the stored constant voltage charging time-current data and normalized.

[0013] S7, determine the actual constant voltage charging data length l d,act And the relationship between the length of the constant voltage charging data stored in step S4, and the corresponding battery actual capacity estimation model for the actual constant voltage charging data can be obtained by querying.

[0014] S8. Substitute the feature variables obtained in step S6 into the battery actual capacity estimation model corresponding to the actual constant voltage charging time obtained in step S7 to estimate the actual capacity value of the battery.

[0015] In step S1, the input terminal of the multi-order network used to characterize the current dynamic characteristics is connected to the positive terminal of the model voltage output terminal; the output terminal of the multi-order network used to characterize the current dynamic characteristics is connected to the negative terminal of the model voltage output terminal; the multi-order network characterizing the current dynamic characteristics includes multiple open-circuit voltage equivalent voltage sources V connected in parallel and in series. eoc,k Resistance R o,k and a resistor-capacitor network; the resistor-capacitor network includes polarization resistors R connected in parallel. p,k and polarization capacitor C p,k Where k = 1, 2, ... n; the analytical mathematical equation of the battery model is:

[0016]

[0017] Where n is the number of parallel branches in the multi-order network characterizing the dynamic characteristics of the current, i.e., the order of the battery model during the constant voltage stage; t is the constant voltage charging time; and I(t) represents the battery constant voltage charging current. k and d k Let I represent the branch current and dynamic characteristic parameter of the k-th order network characterizing current dynamics, respectively. k (0) represents the initial current value of this branch, where d kFor the slope k of the equivalent open-circuit voltage corresponding to this branch ocv,k Sampling frequency f s Resistance R o,k The function of battery capacity Q, i.e.:

[0018] d k =f(k) eocv,k ,f s ,R o,k Q)

[0019] In step S1, determining the battery model structure includes the following steps:

[0020] S1.1, assign the value 1 to parameter k;

[0021] S1.2, Obtain the error e between the estimated and measured current values. I,k And the average execution time T of the parameter identification algorithm com,k ;

[0022] S1.3, Increment the value of parameter k by 1;

[0023] S1.4, determine e I,k-1 -e I,k Value and error difference threshold Δe I And T com,k -T com,k-1 Value and time difference threshold ΔT com Relationship, that is, if e I,k-1 -e I,k <Δe I And T com,k -T com,k-1 >ΔT com If the model order is k-1, then proceed to step S1.2.

[0024] In step S2, the identification process for battery model parameters corresponding to different data lengths includes the following steps:

[0025] S2.1, Extract the constant voltage charging curve of the battery from the previous cycle test as the reference curve, and select the data length corresponding to the current dropping to the cutoff value as the longest data length l required for parameter identification. d,max ;

[0026] S2.2, Set the data length interval Δl d And assign the interval number n to 0;

[0027] S2.3, Identify the corresponding data length l d =l d,max -n*Δl d The battery model parameter set is collected and stored, and the battery model current estimation error e corresponding to the data length is obtained.I,n ;

[0028] S2.4, Increment the interval number n by 1;

[0029] S2.5, Determine the battery model current estimation error e I,n-1 With error threshold e I,max and the number of intervals n and l d,max / Δl d Relationship, that is, if e I,n-1 ≥e I,max or n-1≥l d,max / Δl d Then determine the data length l d,max -(n-1)*Δl d The minimum data length required to identify parameters l d,min Otherwise, proceed to step S2.3.

[0030] In step S3, the expression for the correlation coefficient in the correlation analysis is:

[0031]

[0032] Among them, P θ,Q (l d ) represents the corresponding data length l d The correlation coefficient between the battery model parameter θ and the battery capacity Q, n cycle This refers to the number of battery RPT tests conducted during offline testing.

[0033] In step S3, extract the values ​​that satisfy |P θ,Q (l d The corresponding data lengths of the condition )|>0.8 are used to form the high correlation data length intervals for each model parameter. The number of data points contained in the corresponding high correlation data length intervals for each model parameter is counted. The battery model parameters with a corresponding number of data points greater than a set threshold are selected as the set of feature variables characterizing battery capacity decay.

[0034] In step S3, the normalized expression for the feature variables is:

[0035]

[0036] Among them, FoI i and FoI norm,i These are the feature variable values ​​before and after normalization, and the FoI values ​​are respectively. max and FoI min These are the maximum and minimum values ​​of the feature variables before normalization, respectively.

[0037] In step S4, a battery actual capacity estimation model corresponding to different constant voltage charging data lengths is constructed using the support vector regression algorithm, and the parameters of the support vector regression algorithm are optimized using the gray wolf optimization algorithm to obtain a more accurate battery actual capacity estimation model corresponding to different constant voltage charging data lengths. The method for constructing the battery health state estimation model combining the gray wolf optimization algorithm and the support vector regression algorithm is as follows:

[0038] S4.1, Algorithm parameter initialization: Gray wolf population size N GW Maximum number of iterations T GW ;

[0039] S4.2, Population Initialization: Randomly generate N GW Individual gray wolf x j j = 1, 2, ..., N GW The model is trained using the support vector regression algorithm, and the fitness value of each individual is calculated.

[0040] S4.3 Calculate and record the position vectors x of the leader wolf σ, the second-in-command wolf β, and the advisor wolf δ. σ x β and x δ ;

[0041] S4.4, Update the position of gray wolves in the gray wolf pack;

[0042] S4.5, determine whether the maximum number of iterations T set in S4.1 has been reached. GW Or whether the termination condition is met; if T is not reached. GW If the termination condition is not met, proceed to step S4.2; otherwise, proceed to step S4.6.

[0043] S4.6 outputs the optimal gray wolf position and obtains the optimal parameters for the support vector regression algorithm.

[0044] In step S7, the actual constant voltage charging data length l d,act The process for determining and querying the length relationship of the constant voltage charging data stored in step S4 is as follows: obtain l d,act When l d,act Greater than or equal to l d,max At that time, according to l d,max Select the corresponding battery actual capacity estimation model; otherwise, when l d,act Less than or equal to l d,min At that time, according to l d,min Select the corresponding battery actual capacity estimation model; otherwise, i.e., when l d,act Greater than l d,min And when l d,act Less than l d,max At that time, according to l d,actSelect the corresponding battery actual capacity estimation model.

[0045] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0046] 1. Compared with traditional battery actual capacity estimation methods based on constant voltage conditions, the battery actual capacity estimation method provided by this invention is not sensitive to the preset constant voltage condition cutoff condition, and has high flexibility and robustness.

[0047] 2. The battery equivalent circuit model and structure determination method constructed by this invention for constant voltage conditions have low complexity while ensuring model accuracy;

[0048] 3. The equivalent model parameters selected in this invention can more accurately and effectively characterize the capacity decay of the battery;

[0049] 4. Compared with battery actual capacity estimation methods based on dynamic discharge data and constant current charging data, the estimation method provided by this invention has higher robustness. Attached Figure Description

[0050] Figure 1 This is a flowchart of the battery actual capacity estimation algorithm of the present invention.

[0051] Figure 2 This is a structural diagram of the battery equivalent circuit model for applying the present invention to constant voltage conditions.

[0052] Figure 3 The flowchart shows the process of determining the battery model structure for this invention.

[0053] Figure 4 This is a flowchart illustrating the parameter identification process for battery models with different data lengths according to the present invention.

[0054] Figure 5 A comparison of current estimation errors and parameter identification algorithm running times for different structural models.

[0055] Figure 6 This is a graph showing the evolution trend of the correlation coefficients between the battery model parameters identified based on different constant voltage charging durations and the actual battery capacity.

[0056] Figure 7 Comparison of estimated battery capacity with actual measured values ​​under different constant voltage charging durations. Detailed Implementation

[0057] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0058] This invention provides a method for estimating the actual capacity of lithium-ion batteries under non-perfect constant voltage conditions, enabling efficient and safe management of vehicle-mounted power batteries and energy storage systems.

[0059] A method for estimating the actual capacity of lithium-ion batteries under non-perfect constant voltage conditions is as follows: Figure 1 As shown, this method is mainly divided into two parts: the first part is the offline analysis part, and the second part is the online estimation part; the two parts will be further explained below.

[0060] The offline analysis component includes the following steps:

[0061] 1) Determine the battery model structure based on the test data;

[0062] Battery model structure as follows Figure 2 As shown. The input terminal of the multi-order network used to characterize the current dynamics is connected to the positive terminal of the model's voltage output terminal; the output terminal of the multi-order network used to characterize the current dynamics is connected to the negative terminal of the model's voltage output terminal; the multi-order network characterizing the current dynamics includes multiple open-circuit voltage equivalent voltage sources V connected in parallel and in series. eoc,k Resistance R o,k and a resistor-capacitor network; the resistor-capacitor network includes polarization resistors R connected in parallel. p,k and polarization capacitor C p,k Where k = 1, 2, ... n; the analytical mathematical equation of the battery model is:

[0063]

[0064] Where n is the number of parallel branches in the multi-order network characterizing the dynamic characteristics of the current, i.e., the order of the battery model during the constant voltage stage; t is the constant voltage charging time; and I(t) represents the battery constant voltage charging current. k and d k Let I represent the branch current and dynamic characteristic parameter of the k-th order network characterizing current dynamics, respectively. k (0) represents the initial current value of this branch, where d k For the slope k of the equivalent open-circuit voltage corresponding to this branch ocv,k Sampling frequency f s Resistance Ro,k The function of battery capacity Q, i.e.:

[0065] d k =f(k) eocv,k ,f s ,R o,k Q)

[0066] The process for determining the battery model structure is as follows: Figure 3 As shown, the specific steps include:

[0067] 1.1) Assign the value 1 to parameter k;

[0068] 1.2) Obtain the error e between the estimated and measured current values. I,k And the average execution time T of the parameter identification algorithm com,k ;

[0069] 1.3) Increment the value of parameter k by 1;

[0070] 1.4) Determine e I,k-1 -e I,k Value and error difference threshold Δe I And T com,k -T com,k-1 Value and time difference threshold ΔT com Relationship, that is, if e I,k-1 -e I,k <Δe I And T com,k -T com,k-1 >ΔT com If the model order is k-1, then proceed to step 1.2.

[0071] 2) Identify the battery model parameters corresponding to different data lengths based on the current curve of the constant voltage charging stage of the battery.

[0072] The identification process for battery model parameters corresponding to different data lengths is as follows: Figure 4 As shown, the specific steps include:

[0073] 2.1) Extract the constant voltage charging curve of the battery from the previous cycle test as the reference curve, and select the data length corresponding to the current dropping to the cutoff value as the longest data length l required for parameter identification. d,max ;

[0074] 2.2) Set the data length interval Δl d And assign the interval number n to 0;

[0075] 2.3) Identify the corresponding data length l d =l d,max -n*Δl dThe battery model parameter set is collected and stored, and the battery model current estimation error e corresponding to the data length is obtained. I,n ;

[0076] 2.4) Increment the interval number n by 1;

[0077] 2.5) Determine the current estimation error e of the battery model I,n-1 With error threshold e I,max and the number of intervals n and l d,max / Δl d Relationship, that is, if e I,n-1 ≥e I,max or n-1≥l d,max / Δl d Then determine the data length l d,max -(n-1)*Δl d The minimum data length required to identify parameters l d,min Otherwise, proceed to step 2.3);

[0078] 3) Conduct correlation analysis between battery model parameters and battery capacity for different data lengths throughout the entire life cycle, further determine the effective data length range, and identify and normalize the characteristic variables that characterize battery capacity decay.

[0079] The expression for the correlation coefficient in correlation analysis is as follows:

[0080]

[0081] Among them, P θ,Q (l d ) represents the corresponding data length l d The correlation coefficient between the battery model parameter θ and the battery capacity Q, n cycle This refers to the number of battery RPT tests conducted during offline testing.

[0082] In step 3), extract the values ​​that satisfy |P θ,Q (l d The corresponding data lengths of the condition )|>0.8 are used to form the high correlation data length intervals for each model parameter. The number of data points contained in the corresponding high correlation data length intervals for each model parameter is counted. The battery model parameters with a corresponding number of data points greater than a set threshold are selected as the set of feature variables characterizing battery capacity decay.

[0083] In step 3), the normalized expression for the feature variables is:

[0084]

[0085] Among them, FoI i and FoI norm,iThese are the feature variable values ​​before and after normalization, and the FoI values ​​are respectively. max and FoI min These are the maximum and minimum values ​​of the feature variables before normalization, respectively;

[0086] 4) Combining optimization algorithms and machine learning algorithms, construct and store battery actual capacity estimation models corresponding to different constant voltage charging data lengths;

[0087] Specifically, a battery actual capacity estimation model corresponding to different constant-voltage charging data lengths is constructed using the support vector regression algorithm, and the gray wolf optimization algorithm is used to optimize the parameters of the support vector regression algorithm, resulting in a more accurate battery actual capacity estimation model corresponding to different constant-voltage charging data lengths. The above algorithm can be directly implemented by calling the corresponding algorithm module in MATLAB. The specific implementation steps of the battery health state estimation model construction method combining the gray wolf optimization algorithm and the support vector regression algorithm are as follows:

[0088] S4.1, Algorithm parameter initialization: Gray wolf population size N GW Maximum number of iterations T GW ;

[0089] S4.2, Population Initialization: Randomly generate N GW Individual gray wolf x j j = 1, 2, ..., N GW The model is trained using the support vector regression algorithm, and the fitness value of each individual is calculated.

[0090] S4.3 Calculate and record the position vectors x of the leader wolf σ, the second-in-command wolf β, and the advisor wolf δ. σ x β and x δ ;

[0091] S4.4, Update the position of gray wolves in the gray wolf pack;

[0092] S4.5, determine whether the maximum number of iterations T set in S4.1 has been reached. GW Or whether the termination condition is met; if T is not reached. GW If the termination condition is not met, proceed to step S4.2; otherwise, proceed to step S4.6.

[0093] S4.6 outputs the optimal gray wolf position and obtains the optimal parameters for the support vector regression algorithm.

[0094] The online estimation component includes the following steps:

[0095] 1) Once the battery enters constant voltage charging mode, record and store the battery constant voltage charging time-current sequence;

[0096] 2) After the charging process is completed, the offline analysis steps are identified based on the stored constant voltage charging time-current data.

[0097] 3) The characteristic variables identified are then normalized.

[0098] 3) Determine the actual constant voltage charging data length l d,act And the relationship between the length of the constant voltage charging data stored in the offline analysis step 4) is used to query the battery actual capacity estimation model corresponding to the actual constant voltage charging data.

[0099] Among them, the actual constant voltage charging data length l d,act The process for determining and querying the length relationship of the constant voltage charging data stored in offline analysis step 4) is as follows: obtain l d,act When l d,act Greater than or equal to l d,max At that time, according to l d,max Select the corresponding battery actual capacity estimation model; otherwise, when l d,act Less than or equal to l d,min At that time, according to l d,min Select the corresponding battery actual capacity estimation model; otherwise, i.e., when l d,act Greater than l d,min And when l d,act Less than l d,max At that time, according to l d,act Select the corresponding battery actual capacity estimation model.

[0100] 4) Substitute the feature variables obtained in online estimation step 2) into the battery actual capacity estimation model corresponding to the actual constant voltage charging time obtained in online estimation step 3) to estimate the actual capacity value of the battery.

[0101] In this embodiment, the application is a ternary lithium-ion battery with a nominal capacity of 3.4Ah. However, practical applications are not limited to this. The current estimation error and parameter identification algorithm running time vary depending on the structural model. Figure 5 As shown, the running time of the corresponding parameter identification algorithm increases with the increase of the model order. In comparison, the current estimation accuracy of the second-order model is significantly improved compared to the first-order model; however, the current estimation accuracy of the third-order model is not significantly improved compared to the second-order model. Therefore, the model order is set to 2 for the target battery. The evolution trend of the correlation coefficient between the battery model parameters identified based on different constant voltage charging duration data and the actual battery capacity is shown in the figure. Figure 6 As shown, when the constant voltage charging time is higher than 500s and 590s respectively, the absolute values ​​of the correlation coefficients of the model parameters d1 and I1(0) are both higher than 0.8. Therefore, for the battery to be implemented, d1 and I1(0) are selected as feature variables, and l d,minThe time was set to 590 seconds. Comparison of estimated battery capacity with actual measured values ​​under different constant voltage charging durations obtained using this invention is shown below. Figure 7 As shown, the estimated battery capacity value can be found to track the measured value well. Therefore, this method can effectively achieve online estimation of the actual battery capacity.

[0102] In summary, the present invention provides a method for estimating the actual capacity of lithium-ion batteries under non-complete constant-voltage conditions, comprising: offline determination of the battery model structure; identification of battery model parameters corresponding to different data lengths; further determination of the effective data length range, and simultaneously determination and normalization of characteristic variables representing battery capacity decay; construction and storage of battery actual capacity estimation models corresponding to different constant-voltage charging data lengths; real-time recording and storage of the battery constant-voltage charging time-current sequence; identification and normalization of characteristic variables after the charging process is completed; querying the battery actual capacity estimation model corresponding to the actual constant-voltage charging data; and calculating the actual capacity value of the battery. The estimation method proposed in this invention has the following four advantages: (1) Compared with the traditional battery actual capacity estimation method based on constant voltage conditions, the battery actual capacity estimation method provided by this invention is not sensitive to the preset constant voltage condition cutoff condition and has high flexibility and robustness; (2) The battery equivalent circuit model and structure determination method constructed by this invention for constant voltage conditions has low complexity while ensuring model accuracy; (3) The equivalent model parameters selected by this invention can more accurately and effectively characterize the battery capacity decay; (4) Compared with the battery actual capacity estimation method based on dynamic discharge data and constant current charging data, the estimation method provided by this invention has high robustness.

[0103] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0104] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0105] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for estimating the actual capacity of a lithium-ion battery applied to a nonholonomic constant-voltage operating condition, characterized in that, The method comprises the following steps: S1, determining a battery model structure according to test data; S2, identifying battery model parameters corresponding to different data lengths according to a constant voltage charging stage current curve of the battery; S3, performing correlation analysis of each battery model parameter corresponding to different data lengths and battery capacity in a full life cycle, further determining an effective data length interval, simultaneously determining a characteristic variable representing battery capacity attenuation and normalizing; S4, combining an optimization algorithm and a machine learning algorithm to construct and store battery actual capacity estimation models corresponding to different constant voltage charging data lengths; S5, when the battery enters a constant voltage charging mode, recording and storing a constant voltage charging time-current sequence of the battery; S6, when the charging process ends, identifying the characteristic variable determined in step S3 according to the stored constant voltage charging time-current data and performing normalization conversion; S7, judging the relationship between the actual constant voltage charging data length I d,act and the constant voltage charging data length stored in step S4, querying the battery actual capacity estimation model corresponding to the actual constant voltage charging data; S8, substituting the characteristic variable obtained in step S6 into the battery actual capacity estimation model corresponding to the actual constant voltage charging time obtained in step S7 to estimate the actual capacity value of the battery.

2. A method according to claim 1, characterized in that The input end of the multi-stage network for representing the current dynamic characteristic is connected with the positive pole of the model end voltage output end in the battery model structure determined in the step S1; the output end of the multi-stage network for representing the current dynamic characteristic is connected with the negative pole of the model end voltage output end; the multi-stage network for representing the current dynamic characteristic comprises a plurality of open-circuit voltage equivalent voltage sources V eoc,k , resistors R o,k and resistor-capacitor networks connected in series and in parallel. The resistance-capacitance network comprises a polarization resistance R p,k and a polarization capacitance C p,k wherein k = 1, 2,... n; and the mathematical equation for the battery model analysis is: where n is the number of parallel branches in the multi-order network representing the dynamic characteristics of the current, i.e., the order of the battery model in the constant-voltage stage; t is the constant-voltage charging time, I(t) represents the battery constant-voltage charging current, I k and d k respectively represent the branch current and dynamic characteristic parameter of the kth-order network representing the dynamic characteristics of the current, and I k (0) is the initial current value of the branch, where d k is a function of the equivalent open-circuit voltage slope k ocv,k , the sampling frequency f s , the resistance R o,k , and the battery capacity Q, i.e.: d k = f(k eocv,k ,f s ,R o,k ,Q).

3. A method according to claim 1, characterized in that In the step S1, the determination of the battery model structure comprises the following steps: S1.1, assigning the parameter k to 1; S1.2, obtain error e between current estimation and measurement I,k and parameter identification algorithm average execution time T com,k ; S1.3, increasing the parameter k value by 1; S1.4, judge e I,k-1 -e I,k value and error difference threshold Δe I and T com,k -T com,k-1 value and time difference threshold ΔT com relationship, that is, if e I,k-1 -e I,k <Δe I and T com,k -T com,k-1 >ΔT com , determine the model order k-1, otherwise execute step S1.

2.

4. The method of claim 1, wherein, In the step S2, the identification process of the battery model parameters corresponding to different data lengths comprises the following steps: S2.1, extract the constant voltage charging curve of the battery in the early cycle test as a reference curve, and select the data length corresponding to the current falling to the cutoff value as the longest data length l required for identification parameters d,max ; S2.2, set data length interval Δl d and assign interval number n the value 0; S2.3, identify the corresponding data length l d = l d,max -n*Δl d The battery model parameter set is stored, and the battery model current estimation error e of the corresponding data length is obtained I,n ; S2.4, increasing the interval number n value by 1; S2.5, judging the battery model current estimation error e I,n-1 with the error threshold e I,max and the interval number n and l d,max / Δl d , i.e. if e I,n-1 ≥e I,max or n-1≥l d,max / Δl d , then the data length l d,max -(n-1)*Δl d is determined as the shortest data length l d,min required for the identification parameter, otherwise step S2.3 is executed.

5. The method of claim 1, wherein, In the step S3, the expression of the correlation coefficient in the correlation analysis is: Where, θ i and Q i These are the battery model parameters and battery capacity obtained based on the i-th set of cyclic data, respectively, P θ,Q (l d ) represents the corresponding data length l d The correlation coefficient between the battery model parameter θ and the battery capacity Q, n cycle This refers to the number of battery RPT tests conducted during offline testing. The step S3 extracts the corresponding data length satisfying the condition of |P θ,Q (l d )>0.8, constitutes the high correlation data length interval of each model parameter, counts the data points contained in the corresponding high correlation data length interval of each model parameter, and selects the battery model parameter with a corresponding data point number greater than a set threshold as a characteristic variable set representing the capacity attenuation of the battery. In the step S3, the normalization expression of the characteristic variable is: wherein FoI i and FoI norm,i are the normalized feature variable values, respectively, and FoI max and FoI min are the maximum and minimum values of the feature variable before normalization, respectively.

6. The method of claim 1, wherein, In the step S4, support vector regression is used to construct battery actual capacity estimation models corresponding to different constant voltage charging data lengths, and a grey wolf optimization algorithm is used to optimize support vector regression algorithm parameters to obtain relatively accurate battery actual capacity estimation models corresponding to different constant voltage charging data lengths.

7. The method of claim 1, wherein , the step S7, the actual constant voltage charging data length l d,act And the judgment and query process of the constant voltage charging data length relationship stored in step S4 is: obtaining l d,act When l d,act is greater than or equal to l d,max , the corresponding battery actual capacity estimation model is selected according to l d,max ; otherwise, when l d,act is less than or equal to l d,min , the corresponding battery actual capacity estimation model is selected according to l d,min ; otherwise, that is, when l d,act is greater than l d,min and when l d,act is less than l d,max , the corresponding battery actual capacity estimation model is selected according to l d,act .

8. A method according to claim 6, characterised in that The battery state of health estimation model construction method combining the grey wolf optimization algorithm and the support vector regression algorithm is: S4.1, algorithm parameter initialization: gray wolf population size N GW , maximum number of iterations T GW ; S4.2, population initialization: randomly generate N GW gray wolf individuals x j , j = 1, 2,..., N GW ; train the model by support vector regression algorithm, calculate the fitness value of each individual; S4.3, calculate and record the position vectors x of the alpha wolf, beta wolf and delta wolf σ , x β and x δ ; S4.4, updating the position of the grey wolf in the grey wolf group; S4.5, determine whether the maximum number of iterations T set in S4.1 has been reached. GW Or whether the termination condition is met; if T is not reached. GW And since the termination condition is not met, proceed to step S4.2; Otherwise, step S4.6 is performed; S4.6, outputting the optimal grey wolf position and obtaining optimal support vector regression algorithm parameters.

Images