Equivalent circuit model parameter identification method, device, terminal and storage medium based on EIS

Abstract

The present application discloses an EIS-based equivalent circuit model parameter identification method, device, terminal, and storage medium. The method includes obtaining EIS test data of a battery, screening EIS data corresponding to a preset target SOC, and assigning dynamic weights to EIS data at different frequency points; obtaining a preset equivalent circuit model, setting a parameter range for each model parameter in the equivalent circuit model, and randomly generating an initial population based on a logarithmic distribution within the parameter range; iteratively optimizing the initial population through a differential evolution algorithm, and dynamically adjusting the crossover probability and scaling factor during the iteration process through an adaptive strategy; and outputting the optimal parameters obtained after the iteration when the preset iteration termination condition is met. The method of the present application utilizes a differential evolution algorithm to identify equivalent circuit model parameters, thereby improving the efficiency and accuracy of parameter identification.

Description

Technical Field

[0001] The present application relates to the field of lithium batteries, and in particular to an EIS-based equivalent circuit model parameter identification method, device, terminal, and storage medium. Background Art

[0002] Battery management systems play a key role in ensuring the safe and efficient operation of batteries. The equivalent circuit model parameter identification method based on electrochemical impedance spectroscopy (EIS) can provide important support for battery aging mode analysis, state of health (SOH) estimation and other functions, helping to improve the performance and reliability of battery management systems, thereby promoting the further development of the entire battery application field.

[0003] Currently, traditional EIS-based equivalent circuit model parameter identification methods face many challenges in practical applications. For example, directly selecting EIS data under a specific state of charge (SOC) without considering the differences in data at different frequency points and treating data at all frequency points uniformly will make the final parameter identification results susceptible to the influence of low-quality data, reducing the identification accuracy. Alternatively, for the parameter setting of the equivalent circuit model, only a rough range is given based on experience, and the model parameter range is not precisely set, or an inappropriate initial population generation method is used, which will cause the algorithm to fall into a local optimal solution during the iteration process, making it difficult to find the globally optimal parameters. It is also possible that during iterative optimization, fixed crossover probabilities and scaling factors are used, resulting in the inability to adapt to changes in the iterative process and unable to quickly and effectively converge to the optimal solution, thereby affecting the efficiency and accuracy of the entire parameter identification.

[0004] Therefore, there is an urgent need for an EIS-based equivalent circuit model parameter identification method to improve the accuracy and reliability of battery equivalent circuit model parameter identification and provide more accurate data support for various functions of the battery management system. Summary of the Invention

[0005] In order to solve the above-mentioned problems existing in traditional methods and improve the accuracy and reliability of battery equivalent circuit model parameter identification, the present application provides an equivalent circuit model parameter identification method, device, terminal and storage medium based on EIS.

[0006] In the first aspect, the present application provides an EIS-based equivalent circuit model parameter identification method, which adopts the following technical means:

[0007] An EIS-based equivalent circuit model parameter identification method includes the following steps:

[0008] Obtain the battery's EIS test data, filter the EIS data corresponding to a preset target SOC, and assign dynamic weights to the EIS data at different frequency points;

[0009] Obtaining a preset equivalent circuit model, setting a parameter range for each model parameter in the equivalent circuit model, and randomly generating an initial population based on a logarithmic distribution within the parameter range;

[0010] Iteratively optimizing the initial population through a differential evolution algorithm, and dynamically adjusting the crossover probability and scaling factor during the iteration process through an adaptive strategy;

[0011] When the preset iteration termination condition is met, the optimal parameters obtained after iteration are output.

[0012] By adopting the above technical solution, dynamic weights are assigned to EIS data at different frequency points to improve the impact of errors in the EIS data collected by the chip on model parameter identification; by setting the model parameter range and randomly generating the initial population based on the logarithmic distribution, the model parameter identification accuracy is significantly improved; through the adaptive strategy, the crossover probability and scaling factor are dynamically adjusted to make the algorithm converge faster. By combining all operations and outputting the optimal parameters after meeting the iteration termination conditions, the accurate identification of the battery equivalent circuit model parameters can be completed.

[0013] Preferably, the method of acquiring the EIS test data of the battery, screening the EIS data corresponding to the preset target SOC, and assigning dynamic weights to the EIS data at different frequency points specifically includes the following steps:

[0014] The EIS test data at different states of charge are collected by the EIS detection chip mounted on the battery, and the EIS data corresponding to the preset target SOC is screened and pre-processed.

[0015] The frequency range of the EIS data is divided into multiple frequency intervals according to a preset frequency division rule. Each frequency interval has a corresponding basic weight. The measured data quality of each frequency point is obtained, and the basic weight of each frequency point is adjusted according to the measured data quality to obtain a dynamic weight.

[0016] By adopting the above technical solution, the basic weight of each frequency point is adjusted according to the quality of the measured data to obtain a dynamic weight, which can improve the impact of the error in the EIS data collected by the chip on the identification of model parameters, optimize the model's sensitivity to key information and suppress noise interference.

[0017] Preferably, the method of obtaining a preset equivalent circuit model, setting a parameter range of each model parameter in the equivalent circuit model, and randomly generating an initial population based on a logarithmic distribution within the parameter range specifically includes the following steps:

[0018] Acquire a fractional-order equivalent circuit model including a Weber impedance, wherein the fractional-order equivalent circuit model includes at least one constant-phase element and one Weber element,

[0019] The fractional-order equivalent circuit model is:

[0020] ;

[0021] in, is the equivalent resistance, is the equivalent inductance, is the charge transfer resistance, and is the constant phase equivalent component parameter, is a pseudo capacitor, is the fractional exponential, 、 as well as is the parameter of the Weber element;

[0022] Determining a parameter range of each model parameter in the equivalent circuit model according to the physical characteristics of the battery;

[0023] Within the parameter range, an initial population is randomly generated by logarithmic distribution, each of which contains the model parameters to be identified, specifically,

[0024] A corresponding logarithmic space is generated based on the parameter range of each model parameter, and random numbers corresponding to the model parameters are uniformly generated in each logarithmic space until an initial population of a preset size is generated, where the preset size of the initial population is 1000, and the parameter dimension of the model parameter to be identified in each individual is 8.

[0025] By adopting the above technical solution, by obtaining a fractional-order equivalent circuit model containing specific components and determining the model parameter range according to the physical characteristics of the battery, it is possible to reduce the invalid search space, improve the efficiency and accuracy of the algorithm, and significantly improve the accuracy of model parameter identification; by randomly generating an initial population of a preset size and a parameter dimension of 8 based on a logarithmic distribution within the parameter range, it is possible to make some population parameters randomly generated obey the logarithmic distribution, which helps to improve the effect of model fitting.

[0026] Preferably, the iterative optimization of the initial population by a differential evolution algorithm and the dynamic adjustment of the crossover probability and the scaling factor in the iterative process by an adaptive strategy specifically include the following steps:

[0027] Randomly select three different individuals from the currently generated initial population, , and calculate the mutation vector, i.e.,

[0028] ;

[0029] Where k is the number of iterations and F is the scaling factor;

[0030] The randomly selected target individual and the mutation vector are crossed according to the crossover probability to generate the test individual, that is,

[0031] when hour, ;

[0032] otherwise, ;

[0033] in, is a random number, CR is the crossover probability, is the index of the parameter dimension, is a random integer, For the target individual;

[0034] The fitness of the target individual and the test individual are calculated respectively according to a preset fitness function, wherein the fitness function is:

[0035] ;

[0036] in, is the dynamic weight of each of the frequency points, is the model prediction value, is the measured value;

[0037] If the fitness of the test individual is less than the fitness of the target individual, set k=k+1 and proceed to the next iteration; otherwise, retain the target individual.

[0038] By adopting the above technical solution, the initial population is iteratively optimized through the mutation, crossover, fitness calculation and selection operations in the differential evolution algorithm, gradually approaching the optimal solution; by combining the fitness function with the assigned dynamic weight, the optimization effect can be improved, and the efficiency and accuracy of model parameter identification can be improved, thereby better realizing the equivalent circuit model parameter identification based on EIS.

[0039] Preferably, the method further comprises the following steps:

[0040] The EIS-based equivalent circuit model parameter identification method according to claim 4, further comprising the following steps:

[0041] Obtain the preset initial crossover probability and initial scaling factor, and dynamically update the crossover probability and scaling factor. The formula is:

[0042] ;

[0043] ;

[0044] in, is the initial crossover probability, is the initial scaling factor, K is the total number of iterations, is the preset lower crossover limit.

[0045] By adopting the above technical solution and dynamically updating the crossover probability and scaling factor, the differential evolution algorithm can adaptively adjust these two key parameters during the iteration process, making the algorithm converge faster and thus improving the efficiency of equivalent circuit model parameter identification.

[0046] Preferably, when a preset iteration termination condition is met, outputting the optimal parameters obtained after iteration specifically includes the following steps:

[0047] During the iteration process, the individual with the best fitness in the global iteration process is stored and recorded as the global optimal individual. The model parameters corresponding to the global optimal individual are the optimal parameters. After each iteration generates a new population, the fitness of the optimal individual of the current population is compared with the fitness of the global optimal individual, and the better fitness is taken to update the population.

[0048] After each iteration, check whether the current number of iterations exceeds a preset iteration threshold, or whether the fitness of the current global optimal individual is less than a preset fitness threshold. If any of the above conditions are met, terminate the iteration; otherwise, continue the iteration and determine whether to terminate the iteration based on the preset iteration termination condition.

[0049] After the iteration is terminated, parameter verification and correction are performed based on the optimal parameters and the parameter range, and the corrected optimal parameters are output.

[0050] By adopting the above technical solution, the current optimal parameters are always retained through continuous updates during the iteration process. Whether to terminate the iteration is determined based on the number of iterations and the fitness threshold. The iteration can be stopped in time when the expected effect or the specified number of iterations is achieved to avoid invalid calculations. The optimal parameters are verified, corrected and output based on the parameter range to ensure that the output parameters are consistent with the physical characteristics of the battery, thereby improving the accuracy of the equivalent circuit model parameter identification.

[0051] Preferably, after outputting the optimal parameters, the method further includes the following steps:

[0052] Within the parameter range of each model parameter, a corresponding EIS curve is drawn in combination with the equivalent circuit model, and the EIS curve is analyzed to obtain an intuitive mapping relationship between parameters and graphical features;

[0053] The battery is health assessed based on the comparison of the optimal parameters and the intuitive mapping relationship, and a health assessment report of the battery is generated and output based on the result of the health assessment.

[0054] By adopting the above technical solution, the parameters of the equivalent circuit model are analyzed, and the corresponding EIS curve is drawn in combination with the equivalent circuit model to obtain an intuitive mapping relationship between parameters and graphical features. This mapping relationship and the optimal parameters are then used to evaluate the health of the battery, and a battery health assessment report is generated and output, which is conducive to the subsequent development of application scenarios based on EIS data.

[0055] In a second aspect, the present application provides an EIS-based equivalent circuit model parameter identification device, which adopts the following technical means:

[0056] An EIS-based equivalent circuit model parameter identification device includes the following modules:

[0057] A data preprocessing module is used to obtain the EIS test data of the battery, filter the EIS data corresponding to the preset target SOC, and assign dynamic weights to the EIS data at different frequency points;

[0058] A population initialization module is used to obtain a preset equivalent circuit model, set a parameter range for each model parameter in the equivalent circuit model, and randomly generate an initial population based on a logarithmic distribution within the parameter range;

[0059] an iterative optimization module, configured to iteratively optimize the initial population using a differential evolution algorithm, and dynamically adjust the crossover probability and scaling factor during the iteration process using an adaptive strategy;

[0060] The optimal parameter output module is used to output the optimal parameters obtained after iteration when the preset iteration termination condition is met.

[0061] By adopting the above technical solutions, a parameter identification system based on the equivalent circuit model was built, which provides the necessary software technical support for the accurate and efficient identification of battery parameters and meets the requirements of technological progress in battery management.

[0062] In a third aspect, the present application provides a smart terminal that adopts the following technical solution:

[0063] A smart terminal includes a memory and a processor, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to implement the EIS-based equivalent circuit model parameter identification method as described above.

[0064] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution:

[0065] A computer-readable storage medium stores at least one instruction, at least one program, code set, or instruction set, which is loaded and executed by a processor to implement the EIS-based equivalent circuit model parameter identification method as described above.

[0066] In summary, this application has at least the following beneficial effects:

[0067] (1) By assigning dynamic weights to EIS data at different frequency points, this application can improve the impact of errors in chip-collected EIS data on model parameter identification and improve parameter identification accuracy;

[0068] (2) This application can significantly improve the accuracy of model parameter identification and avoid the algorithm from falling into a local optimal solution by setting the parameter range of each model parameter in the equivalent circuit model and randomly generating the initial population based on the logarithmic distribution within the parameter range;

[0069] (3) This application uses an adaptive strategy to dynamically adjust the crossover probability and scaling factor during the iteration process, which can make the algorithm converge faster, improve the efficiency and accuracy of parameter identification, and realize the use of differential evolution algorithm to identify the parameters of the fractional-order equivalent circuit model. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 is a flow chart of an equivalent circuit model parameter identification method based on EIS according to an embodiment;

[0071] Figure 2 4 is a graph of EIS test data under different SOC states of an equivalent circuit model parameter identification method based on EIS in an embodiment;

[0072] Figure 3 1 is a schematic diagram of parameter ranges of various model parameters in the EIS-based equivalent circuit model parameter identification method according to an embodiment;

[0073] Figure 4 1 is a schematic diagram of fitness corresponding to different crossover probabilities and scaling factors of an equivalent circuit model parameter identification method based on EIS in an embodiment;

[0074] Figure 5 This is one of the corresponding relationship diagrams of the fitness and the number of iterations of the EIS-based equivalent circuit model parameter identification method in the embodiment;

[0075] Figure 6 This is one of the curve diagrams of the model prediction value and the measured value of the equivalent circuit model parameter identification method based on EIS in the embodiment;

[0076] Figure 7 This is a second diagram showing the corresponding relationship between the fitness and the number of iterations of the EIS-based equivalent circuit model parameter identification method in the embodiment;

[0077] Figure 8 This is a second graph of model prediction values ​​and measured values ​​of the EIS-based equivalent circuit model parameter identification method in an embodiment;

[0078] Figure 9 1 is a schematic diagram of various model parameter values ​​of an equivalent circuit model parameter identification method based on EIS in an embodiment;

[0079] Figure 10 is an EIS curve corresponding to the equivalent resistance R1 of the EIS-based equivalent circuit model parameter identification method of the embodiment;

[0080] Figure 11 is an EIS curve corresponding to the equivalent inductance L1 of the EIS-based equivalent circuit model parameter identification method of the embodiment;

[0081] Figure 12 1 is an EIS curve corresponding to the charge transfer resistor R2 of the EIS-based equivalent circuit model parameter identification method of the embodiment;

[0082] Figure 13 1 is an EIS curve corresponding to a constant phase equivalent component parameter Q of an equivalent circuit model parameter identification method based on EIS in an embodiment;

[0083] Figure 14 The constant phase equivalent component parameters of the equivalent circuit model parameter identification method based on EIS are shown in FIG. Corresponding EIS curve graph;

[0084] Figure 15 is the Weber element parameter W of the equivalent circuit model parameter identification method based on EIS in the embodiment R Corresponding EIS curve graph;

[0085] Figure 16 is the Weber element parameter W of the equivalent circuit model parameter identification method based on EIS in the embodiment T Corresponding EIS curve graph;

[0086] Figure 17 is the Weber element parameter W of the equivalent circuit model parameter identification method based on EIS in the embodiment P Corresponding EIS curve graph;

[0087] Figure 18 4 is a structural diagram of an equivalent circuit model parameter identification device based on EIS in an embodiment. DETAILED DESCRIPTION

[0088] The following, in conjunction with the accompanying drawings, provides a clear and complete description of the technical solutions in an embodiment of a key-pair-based channel service request processing method of the present invention. The described embodiments are merely possible technical implementations of the present invention and are not exhaustive. Those skilled in the art can readily devise other embodiments by combining the embodiments of the present invention without inventive effort, and such embodiments are also within the scope of protection of the present invention.

[0089] The EIS-based equivalent circuit model parameter identification method provided in the embodiment of the present application is as follows: Figure 1 As shown, the following steps are included:

[0090] S1. Obtain the battery's EIS test data, filter the EIS data corresponding to the preset target SOC, and assign dynamic weights to the EIS data at different frequency points. This specifically includes the following steps:

[0091] S11. Collect EIS test data at different states of charge using an EIS detection chip mounted on the battery, and filter to obtain EIS data corresponding to a preset target SOC.

[0092] In this embodiment, the EIS detection chip consists of 15 strings of 100Ah battery cells and an NXP chip. The collected EIS test data is also collected based on the NXP chip. The battery is obtained by mounting the EIS detection chip on a 280Ah high-voltage standard box.

[0093] In this embodiment, the EIS test data collected by the NXP chip is as follows: Figure 2 As shown, in this embodiment, the target SOC is set to 90% SOC, so the data of 90% SOC is selected as the AC impedance data for equivalent circuit model parameter identification.

[0094] S12. Preprocessing the collected EIS data.

[0095] Preprocessing includes eliminating abnormal frequency points, such as outliers caused by chip acquisition noise.

[0096] S13. Divide the frequency range of the EIS data into multiple frequency intervals according to a preset frequency division rule. In this embodiment, the frequency intervals include a high frequency band, a mid frequency band, and a low frequency band. Each frequency interval has a corresponding basic weight.

[0097] In this implementation, the frequency division rule is: the frequency of 0.01Hz~1Hz is divided into the low frequency band, and the corresponding basic weight is 1; the frequency of 1Hz~100Hz is divided into the medium frequency band, and the corresponding basic weight is 0.8; the frequency above 100Hz is divided into the high frequency band, and the corresponding basic weight is 0.5.

[0098] Data in high-frequency bands are easily affected by measurement noise, while data in low-frequency bands take longer to measure and may have a lower signal-to-noise ratio. Therefore, the weight of high-frequency band data needs to be reduced due to the large noise, while the weight of low-frequency band data needs to be increased due to the critical information. The weight value can be adjusted.

[0099] S14. Obtain the measured data quality of each frequency point, and adjust the basic weight of each frequency point according to the measured data quality to obtain a dynamic weight.

[0100] The quality of the measured data in this embodiment is the signal-to-noise ratio. For each frequency point, the signal-to-noise ratio of the measured impedance is calculated, that is:

[0101] ;

[0102] in, is the signal strength of the measured impedance, is the noise standard deviation, which can be estimated through multiple measurements or instrument accuracy.

[0103] The basis weight is related to the signal-to-noise ratio, i.e.,

[0104] ;

[0105] in, is the dynamic weight after adjustment of the current frequency point, is the basic weight of the current frequency point.

[0106] In a specific embodiment, if the signal-to-noise ratio of a frequency point is lower than a preset threshold, such as SNR<2, its weight is further reduced or it is directly eliminated.

[0107] The above steps reasonably distribute the data contribution of each frequency point through frequency segmentation and signal-to-noise ratio evaluation, thereby achieving a balance between noise suppression and information extraction and realizing dynamic adjustment of weights.

[0108] S2. Obtain a preset equivalent circuit model, set a parameter range for each model parameter in the equivalent circuit model, and randomly generate an initial population based on a logarithmic distribution within the parameter range, specifically including the following steps:

[0109] S21, obtaining a fractional-order equivalent circuit model including a Weber impedance, the fractional-order equivalent circuit model including at least one constant phase element and one Weber element,

[0110] In this embodiment, from the perspectives of comprehensive accuracy and computing power, the fractional-order equivalent circuit model is a second-order equivalent circuit model, including an equivalent resistor R1, an equivalent inductor L1, a charge transfer resistor R2, a constant phase equivalent element and a Weber element W.

[0111] The fractional-order equivalent circuit model is:

[0112] ;

[0113] in, is the equivalent resistance, is the equivalent inductance, is the charge transfer resistance, and is the constant phase equivalent component parameter, is a pseudo capacitor, is the fractional exponential, 、 as well as are the parameters of the Weber element.

[0114] S22. Determine the parameter range of each model parameter in the equivalent circuit model according to the physical characteristics of the battery.

[0115] For example, for the equivalent resistance R1, a reasonable range is set according to the internal resistance characteristics of the battery. In this embodiment, the typical internal resistance range of a 280Ah battery is used as a reference.

[0116] For the equivalent inductance L1, a reasonable range is set considering the battery polarization effect and material properties;

[0117] For constant phase equivalent components, a reasonable range is set based on the double layer capacitance characteristics of the battery, such as the capacitance strength and characterize the capacitance phase deviation Set the ideal value.

[0118] In this embodiment, the parameter range of each model parameter is as follows: Figure 3 shown.

[0119] S23. Within the parameter range, an initial population is randomly generated by logarithmic distribution. Each initial population contains the model parameters to be identified, specifically,

[0120] S231. Generate a corresponding logarithmic space based on the parameter range of each model parameter.

[0121] For example, if the range of the model parameter x is [a, b], then the corresponding logarithmic space is,

[0122] ;

[0123] S232, uniformly generate random numbers corresponding to model parameters in each logarithmic space,

[0124] ;

[0125] in, is a uniformly distributed random number.

[0126] S233. Restore the original parameter value through exponential restoration:

[0127] .

[0128] If the generated parameters are out of range, they are directly regenerated randomly until all parameters are within the preset range, thereby ensuring the legitimacy of the initial population and preventing invalid solutions from participating in subsequent calculations.

[0129] S234. Repeat the above steps until an initial population of a preset size is generated. The preset size of the initial population is 1000, that is, 1000 individuals are generated. The parameter dimension of the model parameters to be identified in each individual is 8, that is, each individual includes 8 model parameters to be identified, that is, the 8 model parameters in step S21 above, and each individual in the population represents a complete set of model parameter combinations;

[0130] In the current initial population, the i-th individual is represented by ;

[0131] Where D is the parameter dimension of the model parameters to be identified.

[0132] Limiting the parameter range means limiting the search space, which can avoid exploring areas with no physical meaning, avoid wasting resources, improve efficiency, and ensure that the identification results are consistent with the actual characteristics of the battery.

[0133] S3. Iteratively optimize the initial population using a preset differential evolution algorithm and preset iteration rules, and adjust the crossover probability and scaling factor during the iteration process using an adaptive strategy. Specifically, the steps include:

[0134] S31. Randomly select three different individuals from the currently generated initial population. , two of the individuals The vector difference is scaled with the target individual Perform vector synthesis and calculate the mutation vector, that is,

[0135] ;

[0136] in, is the mutation vector, k is the number of iterations, and F is the scaling factor.

[0137] The above mutation operation uses the differences between individuals to generate new solutions, which can enhance the global search capability.

[0138] S32, cross-recombining the randomly selected target individual and the mutant individual corresponding to the mutation vector according to the crossover probability, exchanging part of the information between the two, and generating a test individual, that is,

[0139] when hour, ;

[0140] otherwise, ;

[0141] in, is a random number, is a random integer, j is the parameter dimension index, is the target individual, CR is the crossover probability;

[0142] The above steps can increase the diversity of the population by recombining the target individuals and mutation vectors through crossover operations.

[0143] In a specific implementation method, before the crossover operation, the parameter similarity between the individuals to be crossed is calculated to make an inbreeding judgment: if the similarity between two individuals exceeds a preset threshold, in this embodiment, if the parameter difference is less than 5%, the crossover is rejected and instead crossed with other individuals with larger differences in the population. This can prevent the population from losing its exploration ability due to excessive convergence, thereby maintaining a balance between global search and local development.

[0144] In one specific implementation, after the crossover operation, the parameters of the newly generated test individuals are checked to see if they exceed the parameter range. If so, the parameters are randomly reset to dynamically correct the illegal parameters, maintain the feasibility of the population, and prevent invalid solutions from interfering with fitness evaluation.

[0145] S33. Calculate the fitness of the target individual and the test individual respectively according to the preset fitness function, and select one from the target individual and the test individual to enter the next generation group according to the size of the fitness. The fitness function is:

[0146] ;

[0147] in, is the dynamic weight of each frequency point, is the model prediction value, which corresponds to the real and imaginary parts of the impedance calculated by the model, is the measured value, which corresponds to the real part and imaginary part of the measured impedance. In the above formula, for the convenience of comparison, the fitness is expanded by 10 12 times.

[0148] S34. If the fitness of the test individual is less than the fitness of the target individual, let k=k+1 and the test individual enters the next iteration, otherwise the target individual is retained, that is,

[0149] when hour, ;

[0150] otherwise ;

[0151] in, is the fitness of the test individual, is the fitness of the target individual.

[0152] The above steps retain the best individuals through selection operations, ensuring the quality of the population.

[0153] like Figure 4 As shown in the figure, the fitness corresponding to different crossover probabilities and scaling factors is shown. When the crossover probability is 0.6 and the scaling factor is 0.6, the fitness is optimal, which is 32717. Figure 4 In the data shown, the fitness is expanded by 10 12 The result after doubling.

[0154] like Figure 5 Shown is the corresponding relationship between fitness and number of iterations. Figure 6 is a curve chart of the model prediction value and the measured value, combined with Figure 5 and Figure 6 , Figure 5 The optimal fitness of 32717 when the number of iterations is 5000 is shown in the figure, which corresponds to a crossover probability of 0.6 and a scaling factor of 0.6. Figure 6 It can be seen that when the optimal fitness is 32717, the fitting effect is not ideal.

[0155] In order to improve the above results, after each iteration, the crossover probability and the initial scaling factor are obtained and the crossover probability and the scaling factor are dynamically updated. The formula is:

[0156] ;

[0157] ;

[0158] in, is the initial crossover probability, is the initial scaling factor, K is the total number of iterations, is the preset lower crossover limit.

[0159] In this embodiment, according to the results before improvement, the preset initial crossover probability is 0.6, and the initial scaling factor is 0.6.

[0160] In a specific implementation method, in order to break the premature convergence of the population, increase its diversity, and avoid falling into the local optimum, a disturbance operation and a random restart operation are added.

[0161] During the iteration process, at every fixed generation, N, M% of individuals in the population are randomly selected with probability P. For these selected individuals, their parameters are randomly regenerated to satisfy the range constraints, or they are replaced with the neighborhood solution of the current best individual. In this implementation, N = 500 generations, P = 10%, and M = 10.

[0162] Parameter range constraints and perturbation operations can jointly ensure that the population always explores within the feasible domain and avoid invalid solutions.

[0163] S4. When the preset iteration termination condition is met, the optimal parameters obtained after the iteration are output, and the equivalent circuit model is subjected to parameter analysis based on the optimal parameters to obtain analysis results, which specifically includes the following steps:

[0164] S41. During the iteration process, the individual with the best fitness in the global iteration process, that is, the individual with the smallest fitness, is stored and recorded as the global optimal individual. The model parameters corresponding to the global optimal individual are the optimal parameters. After each iteration generates a new population, the fitness of the optimal individual of the current population is compared with the fitness of the global optimal individual, and the better fitness pair is taken for update.

[0165] In one specific implementation, when the fitness change rate is less than a preset fitness change threshold, the best individual in the current population is selected. This process should execute a local search algorithm on the best individual before the iteration ends. In this embodiment, the particle swarm optimization algorithm is used. The parameters are finely adjusted within its neighborhood, and the optimal solution obtained from the local search replaces the worst individual in the population. This can compensate for the shortcomings of the differential evolution algorithm in local search capabilities and significantly improve parameter identification accuracy.

[0166] S42. Determine whether to terminate the iteration according to a preset iteration termination condition. The iteration termination condition is, after each iteration, checking whether the current number of iterations exceeds a preset iteration threshold, or whether the fitness of the current global optimal individual is less than a preset fitness threshold;

[0167] If any of the above conditions is met, the iteration is terminated, otherwise the iteration continues;

[0168] In this embodiment, the iteration threshold is 10000 and the fitness threshold is 1000, that is, the termination condition is that the number of iterations is greater than 10000 or the fitness of the current global optimal individual is less than 1000.

[0169] like Figure 7 and Figure 8 As shown, Figure 7 When the iteration reaches 10,000, the fitness is 1667. Figure 8 The equivalent circuit fitting effect is obviously improved.

[0170] like Figure 9Shown are the values ​​of each identified model parameter at this time.

[0171] S43. After the iteration is terminated, the parameters are checked and corrected according to the optimal parameters and the parameter range, and the corrected optimal parameters are output.

[0172] In this embodiment, parameter verification includes verifying whether the value of each model parameter is within the parameter range. If the parameter is out of range, it needs to be corrected to be within the range, for example, corrected to the nearest boundary value.

[0173] In a specific embodiment, the parameters are substituted into the equivalent circuit model to generate a simulated EIS curve, and the parameter identification results of the model under different SOCs are obtained. The average fitness is 1555, which is equivalent to the fitting effect using ZView software.

[0174] S44. Within the parameter range of each model parameter, draw a corresponding EIS curve graph in combination with the equivalent circuit model.

[0175] The specific implementation method is that for each model parameter, when other parameters are fixed, 6 logarithmically equidistant points are taken within the corresponding parameter range to draw the EIS curve.

[0176] The range of each parameter is ;

[0177] Divide the range into 5 equally spaced intervals on a logarithmic scale and generate 6 sampling points: , j=0,1,2,3,4,5;

[0178] For each sampling point, substitute the equivalent circuit model and calculate the real part of the impedance and the imaginary part ;

[0179] If the current parameter changes and other parameters are fixed, 6 curves are generated and the changes in the curves are observed.

[0180] In a specific embodiment, Figures 10-17 In order 、 、 、 、 、 、 as well as Corresponding EIS curve diagram.

[0181] S45. Analyze the EIS curve to obtain an intuitive mapping relationship between parameters and graphical features.

[0182] In a specific embodiment, referring to Figures 10-17, we can get the intuitive mapping relationship between the parameters and graphic features of each model parameter:

[0183] Depend on Figure 10 It is found that the larger R1 is, the EIS curve shifts to the right, that is, the real axis intercept in the high-frequency region shifts to the right as R1 increases, and the overall curve shifts.

[0184] The intuitive connection is: R1 directly controls the real part of the high-frequency impedance, reflecting the resistance of the battery itself. An increase in R1 indicates a decrease in electrolyte conductivity, electrode corrosion, or poor contact.

[0185] Depend on Figure 11 It is found that the larger the inductance L1, the more the left side of the EIS curve is pulled down, that is, the low-frequency region (mid-high frequency transition section) bends downward. The larger the L1, the more obvious the downward pull of the left side of the curve.

[0186] The intuitive connection is that L1 is related to the electrochemical polarization or parasitic inductance of the test system, and will introduce inductive resistance when it increases.

[0187] Depend on Figure 12 It is found that the larger the charge transfer resistance R2 is, the larger the semicircle in the EIS curve is, that is, the radius of the intermediate frequency semicircle expands with the increase of R2, and the right endpoint of the semicircle moves to the right.

[0188] The intuitive connection is: R2 represents the difficulty of charge transfer. When it increases, the electrochemical reaction is hindered, reflecting a decrease in electrode activity, such as thickening of the SEI film and deactivation of the catalyst.

[0189] Depend on Figure 13 It is found that the larger the Q is, the higher the right side of the semicircle in the EIS curve is.

[0190] The intuitive connection is that Q reflects the non-ideality of the capacitor. The smaller the Q, the flatter the semicircle and the weaker the capacitance characteristic, indicating that the double layer capacitor is degraded.

[0191] Depend on Figure 14 get, The larger it is, the larger the semicircle diameter and the imaginary part cutting angle are, and the imaginary part cutting angle is the slope in the low-frequency area.

[0192] The intuitive connection is: Control the frequency dependence of the constant phase equivalent element, The anomaly indicates a change in the porous electrode structure.

[0193] Depend on Figure 15 Get, W R The larger it is, the smaller the angle between the semicircle and the straight line is, and the connection is smoother;

[0194] Depend on Figure 16 Get, W T The bigger it is, the shorter the straight line becomes;

[0195] Depend on Figure 17Get, W P Affects the slope of the Weber impedance;

[0196] The intuitive connection is: W reflects the ion diffusion process, W R Increase or W T The decrease indicates that the diffusion resistance increases, W P Anomalies indicate that the diffusion process is hindered, such as an imbalance in the electrolyte concentration gradient.

[0197] S46. Compare the impact of different parameter changes on the curve and identify the sensitive parameters.

[0198] In this embodiment, the high frequency region is significantly affected by R1, the mid-frequency semicircle is controlled by R2 and Q (double-layer capacitance), and the low-frequency slope is controlled by Q and Weber impedance W. P Determine, the sensitive parameters are R1, R2, Q and W P .

[0199] Sensitivity analysis is performed on the combined parameters to obtain fault diagnosis conclusions.

[0200] High-frequency fault: If R1 exceeds 50% of the initial value, it is judged to have changed significantly and diagnosed as electrolyte drying up or current collector corrosion;

[0201] Medium frequency fault: If R2 increases suddenly and the semicircle radius expands, it indicates that the electrode active material is lost or the interface side reaction increases;

[0202] Low-frequency fault: If Q drops significantly and the slope deviates from -45°, it indicates that the electrode material structure collapses or the capacitance characteristics degrade.

[0203] Compared with the change of a single parameter, in more cases the fault type is determined by combining multiple parameter anomalies.

[0204] In a specific embodiment, if R1 increases while Q decreases, it indicates that electrolyte aging and electrode degradation work together.

[0205] If W P Fluctuations and an increase in R2 indicate abnormal diffusion processes and increased charge transfer resistance.

[0206] S47. Perform a health assessment on the battery by combining the optimal parameters and the intuitive mapping relationship, and generate and output a health assessment report of the battery based on the results of the health assessment.

[0207] In this embodiment, sensitive parameters are extracted from the optimal parameters. Since the changes in sensitive parameters in the curve are more significant, in order to save computing resources, sensitive parameters can be extracted for health assessment. In other specific implementable methods, if the accuracy of the results is to be pursued, health assessment can also be performed based on all model parameters.

[0208] Combined with the parameter range of each sensitive parameter, each sensitive parameter is normalized to obtain the corresponding parameter value and eliminate the dimension effect.

[0209] Obtain the health threshold of each sensitive parameter in the battery, calculate the relative change between the current parameter value of the sensitive parameter and the health threshold, and obtain the degradation factor.

[0210] The decay factors of all sensitive parameters are combined with their respective preset weight coefficients for weighted comprehensive calculation to obtain the comprehensive health index of each sensitive parameter.

[0211] Evaluate the battery's health status based on the comprehensive health index and preset health assessment rules.

[0212] In this embodiment, the health assessment rules classify batteries into four levels: excellent, good, critical, and failed, as shown in the following table:

[0213] Health Level SOH Scope Health Assessment excellent >90% Low internal resistance, high capacitance good 70%-90% Mild aging critical 50%-70% Maintenance or replacement required Failure <50% Loss of function

[0214] The optimal parameter list and corresponding fitness values, battery health status level and degradation factor, and fault diagnosis conclusions are summarized into a health assessment report and output.

[0215] Based on the same inventive concept as above, the embodiment of the present application also discloses an equivalent circuit model parameter identification device based on EIS, such as Figure 18 As shown, it includes the following modules:

[0216] The data preprocessing module is used to obtain the EIS test data of the battery, filter the EIS data corresponding to the preset target SOC, and assign dynamic weights to the EIS data at different frequency points;

[0217] A population initialization module is used to obtain a preset equivalent circuit model, set the parameter range of each model parameter in the equivalent circuit model, and randomly generate an initial population based on a logarithmic distribution within the parameter range;

[0218] Iterative optimization module, which is used to iteratively optimize the initial population through the differential evolution algorithm and dynamically adjust the crossover probability and scaling factor during the iteration process through an adaptive strategy;

[0219] The optimal parameter output module is used to output the optimal parameters obtained after iteration when the preset iteration termination condition is met.

[0220] In a specific implementation scheme, the data preprocessing module includes the following units:

[0221] A first data preprocessing unit is used to collect EIS test data at different states of charge through an EIS detection chip mounted on the battery, filter and obtain EIS data corresponding to a preset target SOC, and preprocess the collected EIS data;

[0222] The second data preprocessing unit is used to divide the frequency range of the EIS data into multiple frequency intervals according to a preset frequency division rule. Each frequency interval has a corresponding basic weight, obtain the measured data quality of each frequency point, and adjust the basic weight of each frequency point according to the measured data quality to obtain a dynamic weight.

[0223] In a specific implementation scheme, the population initialization module includes the following units:

[0224] The first population initialization unit is used to obtain a fractional-order equivalent circuit model including a Weber impedance, where the fractional-order equivalent circuit model includes at least one constant phase element and one Weber element.

[0225] The fractional-order equivalent circuit model is:

[0226] ;

[0227] in, is the equivalent resistance, is the equivalent inductance, is the charge transfer resistance, and is the constant phase equivalent component parameter, is a pseudo capacitor, is the fractional exponential, 、 as well as is the parameter of the Weber element;

[0228] The second population initialization unit is used to determine the parameter range of each model parameter in the equivalent circuit model according to the physical characteristics of the battery;

[0229] The third population initialization unit is used to randomly generate an initial population through logarithmic distribution within the parameter range. Each initial population contains the model parameters to be identified, specifically,

[0230] A corresponding logarithmic space is generated based on the parameter range of each model parameter, and random numbers corresponding to the model parameters are uniformly generated in each logarithmic space until an initial population of a preset size is generated. The preset size of the initial population is 1000, where the parameter dimension of the model parameter to be identified in each individual is 8.

[0231] In a specific implementation scheme, the iterative optimization module includes the following units:

[0232] The first iterative optimization unit is used to randomly select three different individuals from the currently generated initial population. , and calculate the mutation vector, i.e.,

[0233] ;

[0234] Where k is the number of iterations and F is the scaling factor;

[0235] The second iterative optimization unit is used to cross the randomly selected target individuals and mutation vectors according to the crossover probability to generate test individuals, that is,

[0236] when hour, ;

[0237] otherwise, ;

[0238] in, is a random number, CR is the crossover probability, is the index of the parameter dimension, is a random integer, For the target individual;

[0239] The third iterative optimization unit is used to calculate the fitness of the target individual and the test individual respectively according to the preset fitness function. The fitness function is:

[0240] ;

[0241] in, is the dynamic weight of each frequency point, is the model prediction value, is the measured value;

[0242] If the fitness of the test individual is less than the fitness of the target individual, set k=k+1 and proceed to the next iteration; otherwise, retain the target individual.

[0243] If the fitness of the test individual is less than the fitness of the target individual, set k=k+1 and proceed to the next iteration; otherwise, retain the target individual.

[0244] The fourth iterative optimization unit is used to obtain the preset initial crossover probability and initial scaling factor, and dynamically update the crossover probability and scaling factor. The formula is:

[0245] ;

[0246] ;

[0247] in, is the initial crossover probability, is the initial scaling factor, K is the total number of iterations, is the preset lower crossover limit.

[0248] In a specific implementation scheme, the optimal parameter output module includes the following units:

[0249] The first optimal parameter output unit is used to store the individual with the best fitness in the global iteration process during the iteration process, which is recorded as the global optimal individual. The model parameters corresponding to the global optimal individual are the optimal parameters. After each iteration generates a new population, the fitness of the optimal individual in the current population is compared with the fitness of the global optimal individual, and the better fitness is selected for update;

[0250] The second optimal parameter output unit is used to check whether the current number of iterations exceeds a preset iteration threshold or whether the fitness of the current global optimal individual is less than a preset fitness threshold after each iteration. If any of the above conditions are met, the iteration is terminated; otherwise, the iteration is continued and whether to terminate the iteration is determined according to the preset iteration termination condition;

[0251] The third optimal parameter output unit is used to perform parameter verification and correction based on the optimal parameter and the parameter range after the iteration is terminated, and output the corrected optimal parameter.

[0252] In a specific implementation scheme, the EIS-based equivalent circuit model parameter identification device further includes the following modules:

[0253] The parameter analysis module is used to draw the corresponding EIS curve within the parameter range of each model parameter in combination with the equivalent circuit model, and analyze the EIS curve to obtain an intuitive mapping relationship between parameters and graphical features; compare the optimal parameters and the intuitive mapping relationship to perform battery health assessment, and generate and output a battery health assessment report based on the results of the health assessment.

[0254] Based on the same inventive concept mentioned above, an embodiment of the present application also discloses a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set. The at least one instruction, at least one program, code set or instruction set can be loaded and executed by a processor to implement the EIS-based equivalent circuit model parameter identification method provided in the above method embodiment.

[0255] Also based on the same inventive concept mentioned above, an embodiment of the present application further discloses a computer-readable storage medium, in which at least one instruction, at least one program, code set or instruction set is stored. The at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the EIS-based equivalent circuit model parameter identification method as described above.

[0256] Those skilled in the art will appreciate that all or part of the steps of the above embodiments may be implemented by hardware or by programs instructing related hardware to implement them. The programs may be stored in computer-readable storage media, which may include, for example, various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0257] The above are only optional embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. An EIS-based equivalent circuit model parameter identification method, characterized in that: The steps include: Obtaining battery EIS test data, screening EIS data corresponding to a preset target SOC, and assigning dynamic weights to the EIS data at different frequency points. Specifically, the frequency range of the EIS data is divided into multiple frequency intervals according to a preset frequency division rule. Each frequency interval has a corresponding basic weight, and the signal-to-noise ratio of the measured impedance at each frequency point is obtained; Among them, Z i is the signal strength of the measured impedance, σ i is the noise standard deviation, and the basic weight is related to the signal-to-noise ratio, that is, Among them, W i is the dynamic weight after adjustment of the current frequency point, W base (f i ) is the basic weight of the current frequency point; Obtain a preset equivalent circuit model, set a parameter range for each model parameter in the equivalent circuit model, and randomly generate an initial population based on a logarithmic distribution within the parameter range, specifically: Iteratively optimizing the initial population through a differential evolution algorithm, and dynamically adjusting the crossover probability and scaling factor during the iteration process through an adaptive strategy; When the preset iteration termination condition is met, the optimal parameters obtained after iteration are output; Within the parameter range of each model parameter, a corresponding EIS curve is drawn in combination with the equivalent circuit model, and the EIS curve is analyzed to obtain an intuitive mapping relationship between parameters and graphical features; The battery is health assessed based on the comparison of the optimal parameters and the intuitive mapping relationship, and a health assessment report of the battery is generated and output based on the result of the health assessment.

2. The EIS-based equivalent circuit model parameter identification method according to claim 1, characterized in that: The steps of obtaining the EIS test data of the battery and screening the EIS data corresponding to the preset target SOC specifically include the following steps: The EIS test data at different states of charge are collected by the EIS detection chip mounted on the battery, and the EIS data corresponding to the preset target SOC is screened and preprocessed.

3. The EIS-based equivalent circuit model parameter identification method according to claim 1, characterized in that: The method of obtaining a preset equivalent circuit model, setting a parameter range of each model parameter in the equivalent circuit model, and randomly generating an initial population based on a logarithmic distribution within the parameter range specifically includes the following steps: Acquire a fractional-order equivalent circuit model including a Weber impedance, wherein the fractional-order equivalent circuit model includes at least one constant-phase element and one Weber element, The fractional-order equivalent circuit model is: Among them, R1 is the equivalent resistance, L1 is the equivalent inductance, R2 is the charge transfer resistance, Q and α are the constant phase equivalent component parameters, Q is the pseudo capacitance, α is the fractional order index, W R 、W T and W P is the parameter of the Weber element; Determining a parameter range of each model parameter in the equivalent circuit model according to the physical characteristics of the battery; Within the parameter range, an initial population is randomly generated by logarithmic distribution, each of which contains the model parameters to be identified, specifically, A corresponding logarithmic space is generated based on the parameter range of each model parameter, and random numbers corresponding to the model parameters are uniformly generated in each logarithmic space until an initial population of a preset size is generated, where the preset size of the initial population is 1000, and the parameter dimension of the model parameter to be identified in each individual is 8.

4. The EIS-based equivalent circuit model parameter identification method according to claim 3, characterized in that: The iterative optimization of the initial population by the differential evolution algorithm and the dynamic adjustment of the crossover probability and the scaling factor in the iterative process by the adaptive strategy specifically include the following steps: Randomly select three different individuals from the currently generated initial population, And calculate the mutation vector, that is, Where k is the number of iterations and F is the scaling factor; The randomly selected target individual and the mutation vector are crossed according to the crossover probability to generate the test individual, that is, When rand(0,1)≤CR or j=j rand hour, otherwise, Among them, rand (0,1) is a random number, CR is the crossover probability, j is the index of the parameter dimension, j rand is a random integer, For the target individual; The fitness of the target individual and the test individual are calculated respectively according to a preset fitness function, wherein the fitness function is: Among them, W i is the dynamic weight of each frequency point, Z i '、z i ' is the model prediction value, Z i ”、z i ” is the measured value; If the fitness of the test individual is less than the fitness of the target individual, set k=k+1 and proceed to the next iteration; otherwise, retain the target individual.

5. The EIS-based equivalent circuit model parameter identification method according to claim 4, characterized in that: The following steps are also included: Obtain the preset initial crossover probability and initial scaling factor, and dynamically update the crossover probability and scaling factor. The formula is: Among them, CR0 is the initial cross probability, F0 is the initial scaling factor, K is the total number of iterations, CR min is the preset lower crossover limit.

6. The EIS-based equivalent circuit model parameter identification method according to claim 4, characterized in that: When the preset iteration termination condition is met, the optimal parameters obtained after iteration are output, which specifically includes the following steps: During the iteration process, the individual with the best fitness in the global iteration process is stored and recorded as the global optimal individual. The model parameters corresponding to the global optimal individual are the optimal parameters. After each iteration generates a new population, the fitness of the optimal individual of the current population is compared with the fitness of the global optimal individual, and the better fitness is taken to update the population. After each iteration, check whether the current number of iterations exceeds a preset iteration threshold, or whether the fitness of the current global optimal individual is less than a preset fitness threshold. If any of the above conditions are met, terminate the iteration; otherwise, continue the iteration and determine whether to terminate the iteration based on the preset iteration termination condition. After the iteration is terminated, parameter verification and correction are performed based on the optimal parameters and the parameter range, and the corrected optimal parameters are output.

7. An EIS-based equivalent circuit model parameter identification device, characterized in that: The steps include: A data preprocessing module is used to obtain EIS test data of the battery, filter the EIS data corresponding to a preset target SOC, assign dynamic weights to the EIS data at different frequency points, divide the frequency range of the EIS data into multiple frequency intervals according to a preset frequency division rule, each frequency interval having a corresponding basic weight, and obtain the signal-to-noise ratio of the measured impedance at each frequency point; Among them, Z i is the signal strength of the measured impedance, σ i is the noise standard deviation, and the basic weight is related to the signal-to-noise ratio, that is, Among them, W i is the dynamic weight after adjustment of the current frequency point, W base (f i ) is the basic weight of the current frequency point; A population initialization module is used to obtain a preset equivalent circuit model, set a parameter range for each model parameter in the equivalent circuit model, and randomly generate an initial population based on a logarithmic distribution within the parameter range; an iterative optimization module, configured to iteratively optimize the initial population using a differential evolution algorithm, and dynamically adjust the crossover probability and scaling factor during the iteration process using an adaptive strategy; The optimal parameter output module is used to output the optimal parameters obtained after iteration when the preset iteration termination condition is met; A parameter analysis module is used to draw a corresponding EIS curve graph in combination with the equivalent circuit model within the parameter range of each model parameter, and analyze the EIS curve graph to obtain an intuitive mapping relationship between parameters and graphical features; The battery is health assessed based on the comparison of the optimal parameters and the intuitive mapping relationship, and a health assessment report of the battery is generated and output based on the result of the health assessment.

8. An intelligent terminal, characterized in that: The invention comprises a memory and a processor, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to implement the EIS-based equivalent circuit model parameter identification method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The readable storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to implement the EIS-based equivalent circuit model parameter identification method as described in any one of claims 1 to 6.

Images