A model-data hybrid-driven SOH evaluation method for lithium-ion batteries under variable operating conditions

Abstract

A model-data hybrid-driven method for evaluating the state of health (SOH) of lithium-ion batteries under varying operating conditions is proposed. This method addresses the poor applicability and low accuracy of existing fixed-operating-condition SOH estimation methods. This method: 1. Accelerated battery aging data is collected under operating conditions 1 and 2. 2. Based on the constant-current discharge curve of the battery under operating condition 1, a particle swarm optimization algorithm is used to identify the SP+ model parameters, obtaining degradation trajectories for model parameters within a battery health range of 100% to 80%. 3. Model parameters with high correlation with the battery SOH are selected as health features and input into a BP neural network algorithm to estimate the battery SOH. The data for operating conditions 1 and 2 serve as the training and test sets, respectively. 4. The BPNN model trained on the operating condition 1 data is fed with the SP+ model parameters for operating condition 2, derived from simulations, and outputs the battery SOH under operating condition 2, completing the evaluation. This method can be used in the battery field.

Description

Technical Field

[0001] The present invention relates to a method for evaluating the health status of a lithium-ion battery, and belongs to the technical field of power battery systems. Background Art

[0002] As an important component of electric vehicles, the state of health (SOH) of lithium-ion batteries is closely related to the safety of electric vehicles and their mileage. It is an important indicator for measuring the degree of battery aging and failure and determining whether the battery has reached the end of its life.

[0003] At the same time, the booming new energy vehicle industry will also generate a large number of retired batteries. Strengthening the cascade utilization of power batteries is currently a key task. However, the degradation patterns of retired batteries during cascade utilization are difficult to predict. For retired lithium-ion batteries with available historical data, extracting health characteristics from this massive amount of data and predicting the battery's state of health (SOH) for cascade utilization is a major challenge. Existing battery SOH estimation methods are often based on fixed operating conditions. In practical applications, SOH assessments for retired batteries undergoing cascade utilization have poor applicability and low accuracy. Summary of the Invention

[0004] The present invention aims to solve the technical problems that the existing SOH estimation methods for lithium-ion batteries under fixed operating conditions have poor applicability and low accuracy, and are unable to effectively carry out SOH evaluation for the cascade utilization of retired batteries, and provide a lithium-ion battery SOH evaluation method under variable operating conditions based on a model-data hybrid drive.

[0005] The present invention provides a method for evaluating the SOH of a lithium-ion battery under variable operating conditions based on a model-data hybrid drive, comprising the following steps:

[0006] Step 1: Conduct a lithium-ion battery accelerated aging test under variable operating conditions at 55-65°C. The charge and discharge rates in operating condition one are both twice those in operating condition two. After the battery is aged under operating condition one to a SOH of 80%, it is further aged under operating condition two to a SOH of 70%. Physical information is collected during the accelerated aging test, wherein the physical information includes battery voltage, current, and capacity.

[0007] Step 2: Based on the constant current discharge curve of the battery under working condition 1, the particle swarm optimization algorithm is used to perform simplified electrochemical (SP+) model parameter identification to obtain the degradation trajectory of the SP+ model parameters in the battery health range of 100% to 80%;

[0008] Step 3: Use the Pearson correlation analysis method to select SP+ model parameters with high correlation with battery SOH. These model parameters are used as health features and input into the BP neural network (BPNN) algorithm to estimate the battery SOH. The data of operating condition 1 is used as the training set, and the data of operating condition 2 is used as the test set.

[0009] Step 4: Based on the BPNN model trained with the data from working condition 1, after the input is changed to the SP+model parameters of working condition 2 obtained by simulation, the battery SOH under working condition 2 is output, which is the SOH of the lithium-ion battery under the variable working condition, completing the evaluation of the battery's health status under the variable working condition.

[0010] Furthermore, the parameters in the simplified electrochemical (SP+) model of the lithium-ion battery described in step 2 are c0, P act_n 、P act_p , Q n , Q p , x0, y0, D x 、D y , τ e , τ n 、R ohm , τ p 、P con ;

[0011] Where c0 is the initial lithium ion concentration in the electrolyte, in mol·m -3 ;

[0012] P act_n is the negative electrode reaction polarization coefficient, unit is m -1.5 ·mol 0.5 ·s;

[0013] P act_p is the positive electrode reaction polarization coefficient, unit is m -1.5 ·mol 0.5 ·s;

[0014] Q n is the capacity of the effective active material of the negative electrode, the unit is C;

[0015] Q p The capacity of the positive electrode active material, unit is C;

[0016] x0 initial lithium insertion rate of negative electrode, dimensionless;

[0017] y0 initial lithium insertion rate of positive electrode, dimensionless;

[0018] D x The range of variation of the negative electrode lithium insertion rate, dimensionless;

[0019] D y The range of variation of the positive electrode lithium insertion rate, dimensionless;

[0020] τ e Liquid phase diffusion time constant, in seconds;

[0021] τ nThe negative electrode solid phase diffusion time constant, in s;

[0022] R ohm Ohmic internal resistance, unit is Ω;

[0023] τ p Positive electrode solid phase diffusion time constant, unit is s;

[0024] P con Liquid phase diffusion coefficient, unit is mol·m -3 ·A -1 .

[0025] Furthermore, the SP+ model parameters that need to be identified using the particle swarm optimization algorithm in step 2 are: negative electrode reaction polarization coefficient P act_n , positive electrode reaction polarization coefficient P act_p , the capacity Q of the negative electrode active material n , the capacity Q of the positive electrode active material p , the initial lithium insertion rate of the negative electrode x0, the initial lithium insertion rate of the positive electrode y0, the negative electrode solid phase diffusion time constant τ n 、Ohmic internal resistance R ohm , positive electrode solid phase diffusion time constant τ p and the liquid phase diffusion coefficient P con .

[0026] Furthermore, in step 2, in order to reduce the workload, it is only necessary to perform identification once every 20 to 40 cycles under working condition 1, for a total of 10 identifications, and use the piecewise cubic Hermite interpolation method to simulate the shape of the degradation trajectory to obtain the SP+ model parameters corresponding to each cycle; then, based on the degradation trajectory of the SP+ model parameters under working condition 1, the support vector machine (SVM) algorithm is used to simulate the degradation trajectory of the SP+ model parameters under working condition 2, that is, in the range of 80% to 70% of battery health, and the identified parameters are substituted back into the SP+ model to fit the simulated voltage curve, ensuring that the error between the simulated voltage curve fitted by the identification parameters and the measured voltage curve is no more than 15mV.

[0027] Furthermore, the calculation formula for the error between the simulated voltage curve fitted by the identification parameters and the measured voltage curve is as follows:

[0028]

[0029] Among them, U actual is the measured voltage value (V), U simulated is the simulated voltage value (V), and N is the total discharge time (s).

[0030] Furthermore, the SP+ model parameters with high correlation with the battery SOH are selected after the Pearson correlation analysis in step 3. High correlation means that the Pearson correlation coefficient between the SP+ model parameters and the SOH is greater than 0.9. The calculation formula of the Pearson correlation coefficient of the two variables is as follows:

[0031]

[0032] Here, n is the number of data points; x and y are the values of the two variables.

[0033] Furthermore, the SP+ model parameter with a high correlation with the battery SOH selected after Pearson correlation analysis in step 3 is the capacity Q of the negative electrode active material. n , the capacity Q of the positive electrode active material p , the initial lithium insertion rate of the negative electrode x0 and the initial lithium insertion rate of the positive electrode y0.

[0034] The present invention adopts the particle swarm optimization algorithm to fragmentarily identify the simplified electrochemical (SP+) model parameters of a part of the working condition, and uses the piecewise cubic Hermite interpolation method (PCHIP) to first simulate the degradation trajectory of the model parameters in working condition one. The advantage of PCHIP is that it can retain the shape of the parameter degradation trajectory to the greatest extent. Then, the support vector machine (SVM) algorithm is used to simulate the degradation trajectory of the model parameters in working condition two. This method of obtaining the degradation trajectory of model parameters under different working conditions throughout the life cycle significantly reduces the computational complexity and time cost, improves the parameter identification efficiency, reduces the data demand, and ensures the accuracy of the model; the B The P neural network (BPNN) model, when its input is replaced with the simulated SP+ model parameters for Condition 2, can accurately output the battery SOH under Condition 2. Therefore, the present invention extracts effective health features based solely on historical charge and discharge data from Condition 1. The trained BPNN model can accurately assess the health status of batteries under Condition 2, thereby simulating the degradation patterns of retired batteries during cascade utilization. Furthermore, the present invention has achieved good test results on self-test data sets at different magnifications, with the average absolute error between the predicted SOH values and the experimental SOH values ranging from 0.85% to 1.2%, demonstrating excellent prediction results and good generalization capabilities. This model can be used in the field of battery SOH assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of the SOH evaluation method for lithium-ion batteries under variable operating conditions based on a model-data hybrid drive according to the present invention;

[0036] Figure 2 This is a flow chart of the particle swarm algorithm identifying SP+ model parameters in step 2 of Example 1;

[0037] Figure 3 This is a graph showing the change in initial lithium insertion rate y0 of the positive electrode of LFP-1 battery with SOH in Example 1;

[0038] Figure 4 The capacity Q of the active material of the positive electrode of LFP-1 battery in Example 1 p Graph showing changes with SOH;

[0039] Figure 5 The capacity Q of the effective active material of the negative electrode of LFP-1 battery in Example 1 n Graph showing changes with SOH;

[0040] Figure 6 This is a graph showing the change in initial lithium insertion rate x0 of the negative electrode of LFP-1 battery with SOH in Example 1;

[0041] Figure 7 This is a comparison diagram of the simulated voltage curve and the measured voltage curve of the LFP-1 battery in Example 1;

[0042] Figure 8 This is the SOH estimation result diagram of LFP-1 battery in Example 1;

[0043] Figure 9 This is the SOH estimation result diagram of the LFP-2 battery in Example 2;

[0044] Figure 10 This is the SOH estimation result diagram of the LFP-3 battery in Example 3. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0046] Example 1: A method for evaluating the SOH of a lithium-ion battery under variable operating conditions based on a hybrid model-data drive is performed in the following steps:

[0047] Step 1. Carry out a variable working condition accelerated aging test of LFP-1 lithium-ion battery (3 lithium iron phosphate / graphite lithium-ion batteries with a rated capacity of 1.5Ah) at 60°C. The variable working condition accelerated aging test method of LFP-1 battery is as follows: first, place the battery in a 60°C constant temperature box for 8 hours to ensure that the temperature inside and outside the battery is consistent and reaches thermal equilibrium. After the shelf life is over, charge and discharge under the working condition, use a constant current constant current charging method, and charge the battery at a constant current rate of 1C until the voltage reaches 3.65V, and then maintain 3.65V for continuous constant voltage charging until the charging current drops below 75mA; then, place the battery on hold for 30 minutes, and after the shelf life is over, use a constant current discharge method to start discharging the battery at a rate of 1C, and the discharge cut-off voltage is 2.0V; after the discharge is completed, continue to place the battery on hold. 30min. After the standby, the above charge and discharge mode is used for cycling until the health of the battery decays to 80%. The charge and discharge operating condition is changed to operating condition 2. First, a constant current constant current charging method is used to charge the battery at a constant current rate of 0.5C until the voltage reaches 3.65V. Then, 3.65V is maintained for continuous constant voltage charging until the charging current drops below 75mA. After that, the battery is placed for 30min. After the standby, a constant current discharge method is used to discharge the battery at a rate of 0.5C, and the discharge cut-off voltage is 2.0V. After the discharge, the battery is placed for another 30min. After the standby, the above charge and discharge mode is used for cycling until the health of the battery decays to 70%. The LFP-1 battery is accelerated aging test under different operating conditions is completed. The interval time for collecting battery voltage, current and capacity information is 1s.

[0048] Step 2. For the constant current discharge curve of the battery under working condition 1, the particle swarm optimization algorithm is used to carry out the simplified electrochemical (SP+) model parameter identification. The particle swarm optimization (PSO) algorithm is an optimization algorithm based on swarm intelligence, which can be effectively used for various parameter identification problems, including the parameter identification of battery electrochemical models. The principle of the particle swarm optimization algorithm to identify the parameters of the simplified electrochemical (SP+) model of the battery is as follows: The particle swarm optimization algorithm simulates the foraging behavior of a flock of birds. The particles move in the solution space and continuously update their positions through information exchange with each other to find the optimal solution. Each particle represents a potential solution, and its position is composed of multiple parameters. In the parameter identification scenario, these parameters correspond to the characteristics of the battery electrochemical model. The flow chart of the particle swarm algorithm identifying the SP+ model parameters is as follows: Figure 2 The specific steps are as follows:

[0049] (1) Initialize the particle swarm:

[0050] N particles are randomly generated. In this embodiment, the value of N is 70. The dimension of each particle is the same as the number of model parameters to be identified. The SP+ model parameters to be identified in this embodiment are: negative electrode reaction polarization coefficient P act_n , positive electrode reaction polarization coefficient P act_p , the capacity Q of the negative electrode active material n , the capacity Q of the positive electrode active material p , the initial lithium insertion rate of the negative electrode x0, the initial lithium insertion rate of the positive electrode y0, the negative electrode solid phase diffusion time constant τ n 、Ohmic internal resistance R ohm , positive electrode solid phase diffusion time constant τ p and the liquid phase diffusion coefficient P con , so the dimension of each particle in this embodiment is 10; a position and speed are randomly initialized for each particle, where the values of position and speed are affected by the manually set model parameter range and identification speed range. The parameter range of LFP-1 lithium iron phosphate battery identified in this embodiment and the identification range set by the optimization algorithm parameter identification are shown in Table 1;

[0051] Table 1 SP+ model parameter identification intervals

[0052]

[0053] (2) Define the objective function:

[0054] An objective function is determined to evaluate the fitness of each particle. For the simplified electrochemical model parameters identified in this embodiment, the objective function used is the root mean square error between the simulated voltage fitted by the identification parameters and the measured voltage.

[0055] (3) Evaluate particle fitness:

[0056] For each particle, i.e., the identified model parameters, the simulated voltage is fitted by back-substituting the simplified electrochemical model, and compared with the measured voltage. The fitness value is calculated by the objective function. For this embodiment, the fitness value is calculated as follows:

[0057]

[0058] Among them, U actual is the measured voltage value (V), U simulated is the simulated voltage value (V), and n is the total discharge time (s).

[0059] (4) Update individual optimality and global optimality:

[0060] Each particle maintains an individual optimal position and records the position with the best fitness. In each iteration, the global optimal position is updated, which is the position with the best fitness among all particles.

[0061] (5) Update particle velocity and position:

[0062] Update the particle's velocity and position according to the following formula:

[0063] Speed update:

[0064] v i =ω×v i +c1×r1×(p i -x i )+c2×r2×(g i -x i )

[0065] Among them, ω is the inertia weight, c1 and c2 are individual and group learning factors respectively, r1 and r2 are random numbers, and p i is the optimal position of the individual, g i is the global optimal position, x i is the current position of the particle, v i is the current velocity of the particle.

[0066] Location Updates:

[0067] x i =x i +v i

[0068] (6) Iteration and termination:

[0069] Repeat steps (3) to (5) until the termination condition is met. The termination condition in this embodiment is that the number of iterations reaches 120.

[0070] (7) Parameter output:

[0071] Ultimately, the parameter values corresponding to the global optimal position are output, representing the optimal values of the simplified electrochemical model parameters for the battery to be identified. The particle swarm optimization algorithm simulates swarm behavior in nature to find the optimal parameters for the battery SP+ model. Its advantages lie in its ease of implementation, ability to handle complex multidimensional search spaces, and insensitivity to initial conditions.

[0072] In order to reduce the workload, it is only necessary to identify once every 25 cycles under working condition one, for a total of 10 identifications, to obtain 10 model parameter sets under working condition one, each parameter set includes 10 SP+ model parameters; the piecewise cubic Hermite interpolation method is used to simulate the shape of the degradation trajectory of each model parameter, and the degradation trajectory of the SP+ model parameters in the battery health range of 100% to 80% is obtained. Then, based on the degradation trajectory of the SP+ model parameters under working condition one, the support vector machine (SVM) algorithm is used to simulate the degradation trajectory of the SP+ model parameters under working condition two, that is, the battery health range of 80% to 70%.

[0073] In step 2, the calculation formula for the error between the simulated voltage curve obtained by identifying the parameters and the measured voltage curve is as follows:

[0074]

[0075] Among them, U actual is the measured voltage value, in V; U simulated is the simulated voltage value, in V; N is the total discharge time, in s.

[0076] The identification parameters are substituted back into the SP+ model to fit the simulated voltage curve, ensuring that the error between the simulated voltage curve fitted by the identification parameters and the measured voltage curve is no more than 15mV, thus ensuring the accuracy of the model.

[0077] Step 3. Pearson correlation analysis is a statistical method used to measure the linear correlation between two variables. The core of this method is to calculate the correlation coefficient (Pearson correlation coefficient), whose value range is between -1 and 1. Specifically, the meaning is: 1 indicates a perfect positive correlation, where as one variable increases, the other also increases; -1 indicates a perfect negative correlation, where as one variable increases, the other decreases; 0 indicates no correlation, where there is no linear relationship between the two variables. The calculation formula of the Pearson correlation coefficient (usually represented by r) is as follows:

[0078]

[0079] Wherein: N is the number of samples, x and y are two variables; for this embodiment, x is the SP+ model parameter as a health feature, and y is the battery health state SOH.

[0080] In this step, the various parameter spaces in the simplified electrochemical SP+ model of LFP-1 lithium iron phosphate battery are: initial lithium ion concentration c0 in the electrolyte, negative electrode reaction polarization coefficient P act_n , positive electrode reaction polarization coefficient P act_p , the capacity Q of the negative electrode active material n , the capacity Q of the positive electrode active materialp , the initial lithium insertion rate of the negative electrode x0, the initial lithium insertion rate of the positive electrode y0, and the range of change of the negative electrode lithium insertion rate D x , the range of change of positive electrode lithium insertion rate D y , liquid phase diffusion time constant τ e , negative electrode solid phase diffusion time constant τ n 、Ohmic internal resistance R ohm , positive electrode solid phase diffusion time constant τ p and the liquid phase diffusion coefficient P con .

[0081] Specifically, the mathematical representation of the model is shown in Table 2, and the definitions of parameters and variables in the SP+ model are shown in Table 3 and Table 4, respectively.

[0082] Table 2 Mathematical representation of SP+ model

[0083]

[0084]

[0085] Table 3 SP+ model parameter definitions

[0086]

[0087]

[0088] Table 4 SP+ model variable definitions

[0089]

[0090] Battery open circuit potential U OCV It is only affected by the lithium ion concentration fractions y and x in the positive and negative electrodes. After the battery is formed (after the battery is manufactured, the internal positive and negative electrode materials are activated through a certain charge and discharge method to improve the battery's charge and discharge performance and self-discharge, storage and other comprehensive performance. This process is called formation) and the loss of lithium ions in a single charge and discharge cycle is ignored. The total lithium content in the positive and negative electrodes is fixed and is only transferred from one electrode to the other. The lithium concentration fraction of the electrode at the end of charging is taken as the initial state, corresponding to y0 and x0. The range of change of y and x during discharge is D y and D x , Among them, Q all is the total capacity of the model. It should be noted that, considering the physical meaning of the model parameters themselves and the empirical rules of the actual parameter acquisition results, D x The valid range for is x0 minus 0 to 0.04.

[0091] The calculation method of the lithium ion concentration fraction of the positive and negative electrodes using the total ion amount is:

[0092]

[0093] During the discharge process, the battery U OCV The calculation formula is:

[0094]

[0095] When current flows through the battery, the ohmic polarization potential η ohm , reaction polarization potential η act It will be generated immediately, and the concentration of solid and liquid lithium ions in the battery will begin to change, but the change in concentration is a slow process. Therefore, the corresponding overpotential also changes gradually, which is described by a time constant in the model.

[0096] η con ,η act ,η ohm Denoted as the total polarization voltage U polar :

[0097] U polar =η con +η act+ η ohm

[0098] The Pearson correlation coefficient between the SP+ model parameters and the battery SOH of LFP-1 lithium iron phosphate battery, and the correlation between the model parameters and the battery SOH are shown in Table 5;

[0099] Table 5 Pearson correlation analysis between SP+ model parameters and battery SOH

[0100]

[0101]

[0102] From the Pearson correlation analysis, we can see that the capacity Q of the negative electrode active material is n , the capacity Q of the positive electrode active material p The initial lithium insertion rate x0 of the negative electrode is strongly positively correlated with the battery SOH, and the initial lithium insertion rate y0 of the positive electrode is strongly negatively correlated with the battery SOH. Therefore, after Pearson correlation analysis, these model parameters are selected as health features to input the BP neural network (BPNN) algorithm to estimate the battery SOH. The data of working condition 1 is the training set, and the data of working condition 2 is the test set.

[0103] Step 4: The BPNN model trained based on the data of working condition 1 can accurately output the battery SOH under working condition 2 after the input is changed to the SP+ model parameters of working condition 2 obtained by simulation, completing the battery health status assessment under variable working condition.

[0104] The error of SOH prediction under variable working conditions for the LFP-1 lithium iron phosphate battery in the test set is shown in Table 6.

[0105] Table 6 Errors in SOH prediction under varying operating conditions for the LFP-1 battery test set

[0106]

[0107] From the results in Table 6, it can be seen that the BP neural network (BPNN) model trained with the data from working condition 1 is used to predict the battery SOH under working condition 2. Taking the mean absolute error as the indicator, the accuracy can reach more than 99%. The maximum absolute error of the SOH predicted under variable working conditions is less than 3%, which verifies that this method can be used to predict the SOH of lithium-ion batteries under variable working conditions.

[0108] In step 3 of this embodiment 1, the initial lithium insertion rate y0 of the positive electrode of the LFP-1 battery changes with SOH as shown in the figure. Figure 3 As shown by Figure 3 It can be seen that the initial lithium insertion rate y0 of the battery positive electrode has a strong negative correlation with SOH. Among them, the parameters of operating condition 1 (SOH: 100%~80%) were identified 10 times in total, and then the piecewise cubic Hermite interpolation method was used to fit the parameter degradation trajectory. This is to simulate the degradation trajectory of the parameters in the historical data of retired batteries through a small number of parameter identifications. The parameters of operating condition 2 (SOH: 80%~70%) were predicted using the support vector machine (SVM) algorithm to simulate and predict the parameter degradation trajectory of retired batteries in the future cascade utilization scenario.

[0109] In step 3 of this embodiment 1, the capacity Q of the LFP-1 battery positive electrode active material is p The change of SOH is shown in the figure below: Figure 4 As shown by Figure 4 It can be seen that the capacity Q of the effective active material of the battery positive electrode p It has a strong positive correlation with SOH. Among them, the parameters of operating condition 1 (SOH: 100% to 80%) were identified 10 times in total, and then the piecewise cubic Hermite interpolation method was used to fit the parameter degradation trajectory. This is to simulate the degradation trajectory of the parameters in the historical data of retired batteries through a small number of parameter identifications. The parameters of operating condition 2 (SOH: 80% to 70%) were predicted using the support vector machine (SVM) algorithm to simulate and predict the parameter degradation trajectory of the future cascade utilization scenario of retired batteries.

[0110] In step 3 of this embodiment 1, the capacity Q of the effective active material of the negative electrode of LFP-1 battery is n The change of SOH is shown in the figure below: Figure 5 As shown, Figure 5 It shows the capacity Q of the effective active material of the battery negative electrode nIt has a strong positive correlation with SOH. Among them, the parameters of operating condition 1 (SOH: 100% to 80%) were identified 10 times in total, and then the piecewise cubic Hermite interpolation method was used to fit the parameter degradation trajectory. This is to simulate the degradation trajectory of the parameters in the historical data of retired batteries through a small number of parameter identifications. The parameters of operating condition 2 (SOH: 80% to 70%) were predicted using the support vector machine (SVM) algorithm to simulate and predict the parameter degradation trajectory of the future cascade utilization scenario of retired batteries.

[0111] In step 3 of this embodiment 1, the initial lithium insertion rate x0 of the negative electrode of LFP-1 battery changes with SOH as shown in the figure. Figure 6 As shown, Figure 6 The results show that the initial lithium insertion rate x0 of the battery negative electrode has a strong positive correlation with the SOH. The parameters of operating condition 1 (SOH: 100% to 80%) were identified 10 times in total. After that, the piecewise cubic Hermite interpolation method was used to fit the parameter degradation trajectory. This was used to simulate the degradation trajectory of the parameters in the historical data of retired batteries through a small number of parameter identifications. The parameters of operating condition 2 (SOH: 80% to 70%) were predicted using the support vector machine (SVM) algorithm to simulate and predict the parameter degradation trajectory of retired batteries in the future cascade utilization scenario.

[0112] Figure 7 This is a comparison chart of the simulated voltage curve and the measured voltage curve of the LFP-1 battery in Example 1. Figure 7 The first cycle of the battery parameter identification was demonstrated. The parameter identification effect was measured by the mean absolute error between the simulated voltage curve and the measured voltage curve. The error was 5.9mV, indicating good identification results.

[0113] Figure 8 This is the SOH estimation result diagram of the LFP-1 battery in Example 1. Figure 8 The mapping relationship between model parameters and battery SOH was demonstrated for LFP-1 battery using BP neural network (BPNN) model. The training set was data from working condition 1 (SOH: 100%~80%), and the test set was data from working condition 2 (SOH: 80%~70%). The charge and discharge rate of working condition 1 of LFP-1 battery was 1C, and the charge and discharge rate of working condition 2 was 0.5C. The average absolute error between the predicted SOH value and the experimental SOH value was used to measure the prediction effect of SOH under variable working conditions. The error was 0.8491%, and the prediction effect was good.

[0114] Example 2: The difference between this example and Example 1 is that in step 1, three lithium iron phosphate / graphite lithium-ion batteries with a rated capacity of 1.5Ah are recorded as LFP-2 batteries. The first charging and discharging condition of the LFP-2 batteries is to use a 2C charge and discharge rate to age to a SOH of 80%, and the second charging and discharging condition is to use a 1C charge and discharge rate to age to 70%. The other steps and parameters are the same as those in Example 1.

[0115] Figure 9 This is the SOH estimation result diagram of LFP-2 battery. Figure 9 The mapping relationship between model parameters and battery SOH was demonstrated for LFP-2 batteries using a BP neural network (BPNN) model. The training set consisted of data from operating condition 1 (SOH: 100% to 80%), and the test set consisted of data from operating condition 2 (SOH: 80% to 70%). The charge and discharge rates for operating condition 1 of the LFP-2 battery were 2C, and those for operating condition 2 were 1C. The prediction effect of SOH under varying operating conditions was measured by the mean absolute error between the predicted SOH value and the experimental SOH value, which was 0.8611%, indicating good prediction results.

[0116] Example 3: This example differs from Example 1 in that in step 1, three 1.5Ah lithium iron phosphate / graphite lithium-ion batteries, designated LFP-3 batteries, are used. The LFP-3 batteries are charged and discharged under condition 1 by aging them at a 3C charge / discharge rate to a SOH of 80%. Condition 2 involves aging them at a 1.5C charge / discharge rate to a SOH of 70%. Other steps and parameters are the same as those in Example 1.

[0117] Figure 10 This is the SOH estimation result diagram of LFP-3 battery. Figure 10 It shows the mapping relationship between model parameters and battery SOH established by using BP neural network (BPNN) model for LFP-3 battery. The training set is the data of working condition 1 (SOH: 100%~80%), and the test set is the data of working condition 2 (SOH: 80%~70%). The charge and discharge rate of working condition 1 of LFP-3 battery is 3C, and the charge and discharge rate of working condition 2 is 1.5C. The average absolute error between the predicted SOH value and the experimental SOH value is used to measure the prediction effect of variable working condition SOH. The error is 1.1717%, and the prediction effect is good.

[0118] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely illustrative of the principles and applications of the invention. It should be understood that many modifications may be made to the illustrative embodiments, and that other arrangements may be devised, without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in ways other than those described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be employed in conjunction with other described embodiments.

Claims

1. A lithium-ion battery variable operating condition SOH assessment method based on a hybrid model-data drive, characterized by The method comprises the following steps: Step 1: Conduct a lithium-ion battery accelerated aging test under variable operating conditions at 55-65°C. The charge and discharge rates in operating condition one are both twice those in operating condition two. After the battery is aged under operating condition one to a SOH of 80%, it is further aged under operating condition two to a SOH of 70%. Physical information is collected during the accelerated aging test, including battery voltage, current, and capacity. Step 2. For the constant current discharge curve of the battery under working condition 1, the particle swarm optimization algorithm is used to carry out the simplified electrochemical SP+ model parameter identification. Specifically, under working condition 1, the identification is performed once every 20 to 40 cycles, for a total of 10 times, and the shape of the degradation trajectory is simulated using the piecewise cubic Hermite interpolation method to obtain the SP+ model parameters corresponding to each cycle; then, based on the degradation trajectory of the SP+ model parameters under working condition 1, the support vector machine algorithm is used to simulate the degradation trajectory of the SP+ model parameters under working condition 2, that is, the battery health range of 80% to 70%, and the identification parameters are substituted back into the SP+ model to fit the simulated voltage curve, ensuring that the error between the simulated voltage curve fitted by the identification parameters and the measured voltage curve is no more than 15mV; the degradation trajectory of the SP+ model parameters in the battery health range of 100% to 80% is obtained; Step 3: Use the Pearson correlation analysis method to select SP+ model parameters with high correlation with battery SOH. These model parameters are used as health features to input the BP neural network algorithm to estimate the battery SOH. The data of working condition 1 is used as the training set, and the data of working condition 2 is used as the test set. Step 4: Based on the BPNN model trained with the data from working condition 1, after the input is changed to the SP+model parameters of working condition 2 obtained by simulation, the battery SOH under working condition 2 is output, which is the SOH of the lithium-ion battery under the variable working condition, completing the evaluation of the battery's health status under the variable working condition.

2. The method for evaluating SOH of lithium-ion batteries under variable operating conditions based on model-data hybrid drive according to claim 1, characterized in that: The parameters in the simplified electrochemical SP+ model of the lithium-ion battery described in step 2 are 、 、 、 、 、 、 、 、 、 、 、 、 、 ; in, is the initial lithium ion concentration in the electrolyte, in units of ; is the negative electrode reaction polarization coefficient, unit is ; is the positive electrode reaction polarization coefficient, unit is ; is the capacity of the effective active material of the negative electrode, the unit is C; The capacity of the positive electrode active material, unit is C; Initial lithium insertion rate of negative electrode, dimensionless; Initial lithium insertion rate of the positive electrode, dimensionless; The range of variation of the negative electrode lithium insertion rate, dimensionless; The range of variation of the positive electrode lithium insertion rate, dimensionless; Liquid phase diffusion time constant, in seconds; The negative electrode solid phase diffusion time constant, in s; Ohmic internal resistance, unit is Ω; Positive electrode solid phase diffusion time constant, unit is s; Liquid phase diffusion coefficient, in units of .

3. The method for evaluating SOH of a lithium-ion battery under variable operating conditions based on a model-data hybrid drive according to claim 1 or 2, characterized in that: The SP+ model parameters identified by the particle swarm optimization algorithm in step 2 are: negative electrode reaction polarization coefficient , positive electrode reaction polarization coefficient , the capacity of the negative electrode active material , the capacity of the positive electrode active material , initial lithium insertion rate of negative electrode , initial lithium insertion rate of the positive electrode , negative electrode solid phase diffusion time constant , ohmic internal resistance , positive electrode solid phase diffusion time constant and the liquid phase diffusion coefficient .

4. The method for evaluating SOH of lithium-ion batteries under variable operating conditions based on model-data hybrid drive according to claim 1, characterized in that: In step 2, the calculation formula for the error between the simulated voltage curve obtained by identifying the parameters and the measured voltage curve is as follows: ， in, is the measured voltage value, in V; is the simulated voltage value, in V; is the total discharge time in seconds.

5. The method for evaluating SOH of lithium-ion batteries under variable operating conditions based on model-data hybrid drive according to claim 1 or 2, characterized in that: The SP+ model parameters with high correlation with battery SOH are selected after Pearson correlation analysis in step 3. High correlation means that the Pearson correlation coefficient between the SP+ model parameters and SOH is greater than 0.

9. The calculation formula of the Pearson correlation coefficient of the two variables is as follows: ， in, is the number of data points; are the values of two variables.

6. The method for evaluating SOH of lithium-ion batteries under variable operating conditions based on model-data hybrid drive according to claim 1 or 2, characterized in that: The SP+ model parameter with a high correlation with the battery SOH selected after Pearson correlation analysis in step 3 is the capacity of the negative electrode active material. , the capacity of the positive electrode active material , initial lithium insertion rate of negative electrode and the initial lithium insertion rate of the positive electrode .

Images