Battery electrochemical parameter identification method, system, equipment and program product

Abstract

The invention provides a battery electrochemical parameter identification method, system and device and a program product, and the method comprises the steps: carrying out the discharge test of a plurality of discharge rates on a battery, and obtaining the experimental data of the discharge test; constructing a battery electrochemical model based on experimental data; performing cross-working-condition sensitivity analysis on the model input parameters of the battery electrochemical model by adopting a global sensitivity analysis method to obtain global sensitivity parameters; constructing a target function based on key region constraint based on experimental data; and alternately adopting a constraint Bayesian optimization method based on a trust domain and a granular self-adaptive local search method to explore the optimized target function, and iteratively optimizing the target function to obtain an optimal solution of the global sensitivity parameter. According to the method, a set of parameter identification system with high precision, cross-working-condition robustness and calculation efficiency is constructed, and cross-working-condition high-precision identification of the electrochemical parameters of the high-capacity lithium ion battery is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery parameter identification, and particularly relates to a battery electrochemical parameter identification method, system, device and program product. BACKGROUND

[0002] The physical model parameter identification technology of high-capacity lithium ion batteries in current grid-level energy storage systems faces the problem of systematic adaptation of algorithm framework and engineering needs. In the physical model parameter identification of high-capacity lithium ion batteries (>200 Ah) in grid-level energy storage systems, the existing technology causes the systematic mismatch of algorithm framework and engineering needs, so that the electrochemical parameter identification accuracy and efficiency cannot meet the actual application needs. The traditional method (such as Bayesian optimization, gradient method) cannot effectively coordinate the optimization path of global exploration and local convergence, relies on static statistical model or single working condition optimization strategy, and is easy to fall into local extreme value or divergence under cross-working condition, so that the parameter estimation error is amplified, which seriously restricts the high-precision prediction ability of the battery management system to the key states such as internal lithium concentration and reaction rate, and it is difficult to cope with the multi-physical field strong coupling characteristics and cross-working condition dynamic needs caused by the geometric amplification of large-capacity batteries. At the same time, the strong coupling characteristics of solid-liquid phase diffusion, electrochemical reaction and thermal effect caused by the geometric amplification of large-capacity batteries lead to the high dimension of parameter space and the aggravation of numerical stiffness problem.

[0003] Although the parameter identification method based on Bayesian optimization can partially alleviate the exploration problem of high-dimensional parameter space, its computational complexity increases cubically with the dimension of the covariance matrix, making it difficult to extend to refined models that include solid-liquid phase diffusion, electrochemical reaction and thermal effect coupling. Although the local optimization algorithm based on gradient descent has an advantage in convergence speed, it is frequently unstable due to the ill-conditioned Hessian matrix caused by the numerical stiffness of the model, and the voltage prediction error in the safety sensitive interval such as full charge and deep discharge is significantly higher than that in the normal working condition, directly affecting the real-time warning ability of the battery management system (BMS) on lithium precipitation risk and life attenuation. The existing identification methods (such as meta-heuristic algorithms) are difficult to expand due to the exponential growth of computational complexity, and lack of targeted optimization in safety sensitive intervals such as full charge and deep discharge, resulting in a significant increase in voltage prediction error at the safety boundary, hindering the lithium precipitation warning and life attenuation control in extreme working conditions. In addition, the existing methods generally lack a systematic design for parameter generalization across working conditions. Traditional methods (such as Bayesian optimization and hybrid meta-heuristic algorithms) do not embed cross-condition sensitivity analysis mechanisms, resulting in a disconnection between the parameter identification process and the multi-physical field coupling characteristics in actual battery operation. The difference in time scales between solid-liquid phase diffusion and electrochemical reaction caused by the geometric amplification of high-capacity batteries results in a strong non-linear correlation in the parameter space. Existing global optimization algorithms do not decouple the action paths of key parameters, and when the grid-level energy storage system is actually running, the voltage prediction error increases exponentially when the rate is switched from 0.1C to 1C or the temperature suddenly changes, resulting in a sharp decline in parameter identification accuracy, directly weakening the real-time sensing ability of the battery management system for internal lithium deposition risk, and severely restricting the safety boundary control of high-capacity batteries in complex scenarios such as energy time shifting and frequency modulation auxiliary services.

[0004] At the same time, the existing identification framework generally uses a homogenized objective function design, without differentiating the error constraints in safety sensitive intervals such as full charge and deep discharge, resulting in a prediction deviation of lithium concentration gradient of the model exceeding 10% in extreme working conditions, significantly increasing the probability of misjudgment of thermal runaway. More seriously, traditional methods lack parameter dimension reduction mechanisms and efficient convergence strategies, resulting in a significant increase in computational time when facing high-dimensional mathematical models, with a single full parameter identification period lasting more than 10 hours, which cannot meet the rigid demand of minute-level state update of the grid-level energy storage system. SUMMARY

[0005] To solve the problems in the prior art, the purpose of the present application is to provide a battery electrochemical parameter identification method, system, device and program product, effectively solving the contradiction between the traditional method in the calculation efficiency, generalization ability and safety boundary accuracy, building a set of parameter identification system with high precision, cross-condition robustness and high efficiency, realizing the cross-condition high-precision identification of high-capacity lithium ion battery electrochemical parameters.

[0006] The first aspect of the application provides a battery electrochemical parameter identification method, comprising the following steps:

[0007] Obtaining experimental data obtained by performing discharge tests on the battery at multiple discharge rates;

[0008] Constructing a battery electrochemical model based on the experimental data;

[0009] Performing cross-condition sensitivity analysis on the model input parameters of the battery electrochemical model using a global sensitivity analysis method to obtain global sensitivity parameters;

[0010] Constructing a target function based on key region constraints based on the experimental data;

[0011] Alternately using a trust region-based constrained Bayesian optimization method and a granular adaptive local search method to explore the target function, iteratively optimizing the target function, and obtaining an optimal solution of the global sensitivity parameters.

[0012] In some embodiments, the cross-condition sensitivity analysis on the battery electrochemical model obtains global sensitivity parameters, comprising the following steps:

[0013] Calculating the model output of the battery electrochemical parameters corresponding to various discharge rates, and obtaining multi-objective sensitivity indicators according to the model output at various discharge rates;

[0014] Using a global sensitivity analysis method to quantitatively evaluate the sensitivity index of each model input parameter of the battery electrochemical parameters to the model output indicators, wherein the sensitivity index includes the sensitivity index of a single model input parameter to the voltage response and the sensitivity index of parameter synergy to the voltage response, and the model output indicators are determined according to the multi-objective sensitivity indicators;

[0015] According to the sensitivity index of each model input parameter of the battery electrochemical parameters, the global sensitivity parameters in the model input parameters are screened.

[0016] In some embodiments, calculating the model output of the battery electrochemical parameters corresponding to various discharge rates, and obtaining multi-objective sensitivity indicators according to the model output at various discharge rates, comprises the following steps:

[0017] Determining the model output corresponding to various discharge rates according to the root mean square error between the simulation voltage curve and the experimental voltage curve of the battery electrochemical parameters corresponding to various discharge rates;

[0018] Obtaining multi-objective sensitivity indicators from the model output at various discharge rates and the relative importance weight vector of different discharge rates.

[0019] In some embodiments, constructing a target function based on key region constraints based on the experimental data, comprises the following steps:

[0020] Based on the experimental data, the following formula is used to construct the working condition error function based on the root mean square error as the objective function:

[0021]

[0022] in, represents the working condition error, θ represents the global sensitivity parameter vector to be optimized, Indicates the discharge rate C at the i-th i The experimental data set when N i Indicates the discharge rate C i The number of sampling time points of the experimental data is: The battery electrochemical model based on the parameter θ is expressed as i Time t j The predicted terminal voltage when Indicates the discharge rate C i Time t j The experimental measurement terminal voltage at ;

[0023] The minimax robust objective is constructed as follows:

[0024]

[0025] Where Θ represents the feasible space of the global sensitivity parameter vector to be optimized;

[0026] The following regional constraints are imposed on the objective function based on the minimax robust objective:

[0027]

[0028] in, is the sub-dataset of the high state of charge interval in the experimental data set, represents the sub-dataset of the low state of charge interval in the experimental data set, ε high and ε low They respectively represent the maximum allowable terminal voltage deviation corresponding to the high state of charge interval and the maximum allowable terminal voltage deviation corresponding to the low state of charge interval.

[0029] In some embodiments, a constrained Bayesian optimization method based on a trust region and a granular adaptive local search method are alternately used to explore the optimization objective function, iteratively optimize the objective function, and obtain the optimal solution of the global sensitivity parameter, including performing the following steps in each iteration round:

[0030] The constrained Bayesian optimization method based on the trust region performs a global exploration of the optimization objective function and obtains the global optimal solution of the current iteration round;

[0031] Based on the granular adaptive local search method, the current global optimal solution is adjusted to obtain a local optimal solution, which is used as the basis for constructing the trust region in the next iteration.

[0032] In some embodiments, the trust region based constrained Bayesian optimization method performs global exploration on the optimization objective function to obtain a global optimal solution in the current iteration, including the following steps:

[0033] Gaussian process priors are constructed on the global parameter vector respectively;

[0034] In the first round, Latin hypercube design is used for initial sampling to obtain initial sampling points, and the trust region is determined according to the initial sampling points and the initial trust region radius. In the subsequent rounds, the trust region is determined based on the trust region radius and the local optimal solution updated in the previous round;

[0035] Within the trust region, the candidate optimal solution is obtained by minimizing the expected improvement, and the global optimal solution is updated using the candidate optimal solution, and the trust region radius is updated.

[0036] In some embodiments, based on the granular adaptive local search method, the current global optimal solution is adjusted to obtain a local optimal solution, including the following steps:

[0037] Based on the current global optimal solution, the grid of the search sub-stage is constructed as follows:

[0038]

[0039] where M k represents the search sub-stage grid, θ k represents the current global optimal solution, represents the grid perturbation amplitude of the search sub-stage, and z represents an integer vector, represents a p-dimensional integer space, and p is the number of battery electrochemical parameters to be optimized;

[0040] In the search sub-stage, a finite batch of grid candidates is generated, and if there is an improvement point, the grid is enlarged and the θ k If the search stagnates, enter the finalization sub-stage as follows:

[0041]

[0042] where D k is a positive definite set, is the polarization step size of the finalization sub-stage, and d represents a direction vector in the positive definite set. If there is still no improvement, the grid is contracted and the polarization step size is Stop searching to obtain a local optimal solution.​

[0043] The second aspect of the present application further provides a battery electrochemical parameter identification system for implementing the battery electrochemical parameter identification method of the first aspect, and the system comprises:

[0044] a data acquisition module configured to acquire experimental data obtained by performing discharge tests on the battery at multiple discharge rates;

[0045] a model construction module configured to construct a battery electrochemical model based on the experimental data;

[0046] a sensitivity analysis module configured to perform cross-condition sensitivity analysis on model input parameters of the battery electrochemical model by using a global sensitivity analysis method to obtain global sensitivity parameters;

[0047] a parameter identification module configured to construct a target function based on a key region constraint based on the experimental data, and alternately use a trust region-based constrained Bayesian optimization method and a granularized adaptive local search method to explore and iteratively optimize the target function to obtain an optimal solution of the global sensitivity parameters.

[0048] The third aspect of the present application further provides a battery electrochemical parameter identification device, comprising:

[0049] a processor;

[0050] a memory having executable instructions of the processor stored therein;

[0051] The processor is configured to execute the steps of the battery electrochemical parameter identification method described above by executing the executable instructions.

[0052] The fourth aspect of the present application further provides a computer program product comprising a computer program, which, when executed by a processor, implements the steps of the battery electrochemical parameter identification method described above.

[0053] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application.

[0054] The battery electrochemical parameter identification method, system, device and program product of the present application have the following advantages

[0055] Advantages:

[0056] The present application aims to build a parameter identification system with high precision, cross-condition robustness and high computational efficiency. In the present application, a cross-condition parameter sensitivity dynamic screening mechanism is introduced. In the mechanism, the dimensionality of the driving parameters is reduced through cross-condition sensitivity analysis, the response characteristics of the core parameters under the coupling effect of multiple conditions are dynamically analyzed, and the core parameter set sensitive to the response of multiple conditions is dynamically screened to realize adaptive compression of the model dimensionality. A sequential global-local optimization architecture is designed to organically integrate the global exploration capability of Bayesian optimization and the constraint convergence characteristics of local grid search method. The two-stage Bayesian-local grid search hybrid optimization strategy realizes the cooperation of global exploration and local convergence, solves the convergence contradiction, and solves the mutual exclusion problem of search efficiency in high-dimensional space and local extremum escape. At the same time, an optimization objective function based on regional perception is introduced, and a regional perception minimum-maximum optimization objective is embedded to impose hard constraints on the voltage tracking error in safety-sensitive intervals such as full charge and deep discharge, ensuring that the prediction error of lithium ion concentration under full charge condition is less than 1%, and laying a model foundation for accurate management of the safety boundary of the energy storage system. Finally, through the deep adaptation of the algorithm architecture and engineering requirements, the present application promotes the paradigm shift of high-capacity lithium ion battery parameter identification technology from "offline coarse calibration" to "online high precision". Through the present application, high-precision and cross-condition robust identification of high-capacity lithium ion battery electrochemical parameters is realized, which provides reliable technical support for real-time state estimation and safety management of grid-level energy storage systems, and can greatly improve the condition generalization ability and safety warning reliability of grid-level energy storage battery models, making the parameter identification efficiency of the models in dynamic scenarios such as energy time shift and frequency modulation response increase by more than 5 times, and providing high-precision bottom model support for real-time state estimation and life prediction of kilowatt-level energy storage power stations. BRIEF DESCRIPTION OF DRAWINGS

[0057] Other features, objects, and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0058] Figure 1 is a flowchart of a battery electrochemical parameter identification method according to an embodiment of the present application;

[0059] Figure 2 is a schematic diagram of a DFN electrochemical model;

[0060] Figure 3a is a schematic diagram of the modeling and sensitivity analysis part in the overall framework of the battery parameter identification according to an embodiment of the present application;

[0061] Figure 3b is a schematic diagram of the parameter identification part in the overall framework of the battery parameter identification according to an embodiment of the present application;

[0062] Figure 4is a schematic diagram of Sobol first-order sensitivity index of an embodiment of the present application;

[0063] Figure 5 is a schematic diagram of Sobol total sensitivity index of an embodiment of the present application;

[0064] Figure 6 is a schematic diagram of discharge curve of battery No. 81 under multi-rate discharge of an embodiment of the present application;

[0065] Figure 7 is a schematic diagram of discharge curve of battery No. 82 under multi-rate discharge of an embodiment of the present application;

[0066] Figure 8 is a comparison diagram of discharge results of different parameters under 0.33C of an embodiment of the present application;

[0067] Figure 9 is a comparison diagram of discharge results of different parameters under 1C of an embodiment of the present application;

[0068] Figure 10a is a schematic diagram of dynamic current working condition test of an embodiment of the present application;

[0069] Figure 11 is a structural schematic diagram of a battery electrochemical parameter identification system of an embodiment of the present application;

[0070] Figure 12 is a structural schematic diagram of a battery electrochemical parameter identification device of an embodiment of the present application. DETAILED DESCRIPTION

[0071] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations can be implemented in any

[0072] Moreover, the drawings represent a simplified schematic diagram of the present application and are not necessarily to scale. Identical reference numerals in the figures represent the same or similar parts and their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities that do not necessarily have to correspond to physically or logically independent entities. These functional entities can be implemented in software form or in one or more hardware modules or integrated circuits or in different network and / or processor devices and / or microcontroller devices. Although the present specification uses “first” or “second” or the like to represent certain features, it is merely to represent the role and not as a limitation on the number and importance of the specific features.

[0073] The flowchart shown in the drawing is only an exemplary illustration and does not necessarily include all the steps. For example, some steps can be further decomposed, and some steps can be combined or partially combined, so that the actual execution order can be changed according to the actual situation.

[0074] As Figure 1 shown, the embodiment of the application provides a battery electrochemical parameter identification method, comprising the following steps:

[0075] S100: obtaining experimental data obtained by performing discharge tests on the battery at multiple discharge rates;

[0076] S200: constructing a battery electrochemical model based on the experimental data;

[0077] The basic framework of the lithium ion battery electrochemical parameter estimation adopted by the application is based on the Doyle-Fuller-Newman (DFN) model as the core theoretical basis; the DFN model is the most comprehensive physical and chemical model for describing the internal electrochemical process of the lithium ion battery, as Figure 2 shown, the internal structure of the battery is discretized into three different regions of negative electrode, separator and positive electrode, each electrode is composed of spherical active material particles, lithium ions occur intercalation and deintercalation reactions in them, and the model describes the coupling phenomena of mass transfer, charge transfer and electrochemical reaction through a series of partial differential equations;

[0078] S300: performing cross-condition sensitivity analysis on the model input parameters of the battery electrochemical model by using a global sensitivity analysis method to obtain global sensitivity parameters;

[0079] S400: constructing a target function based on key region constraints based on the experimental data;

[0080] S500: alternately using a trust region-based constrained Bayesian optimization method and a granular adaptive local search method to explore and iteratively optimize the target function, and obtaining the optimal solution of the global sensitivity parameters.

[0081] The fundamental limitation of the current high-capacity lithium ion battery electrochemical parameter identification technology is due to the fragmented optimization path of global search and local convergence, and the lack of cross-condition generalization mechanism and safety boundary constraints, which seriously limits the prediction accuracy and engineering applicability of the model in the dynamic scenario of grid-level energy storage.

[0082] The present application aims to build a set of parameter identification system with high precision, cross-condition robustness and high efficiency of calculation. Through the construction of a sequential optimization architecture driven by cross-condition sensitivity, the contradiction between calculation efficiency, generalization ability and safety boundary precision in existing design is resolved from the causal level. In the present application, a cross-condition parameter sensitivity dynamic screening mechanism is introduced. Through cross-condition sensitivity analysis driven parameter dimension reduction, the response characteristics of core parameters under the coupling effect of multiple working conditions are dynamically analyzed, and the core parameter set sensitive to multiple working condition responses is dynamically screened to realize adaptive compression of model dimension. The degree of freedom of high-dimensional mathematical model is compressed by more than 50%, significantly reducing the covariance matrix calculation dimension in the Bayesian optimization stage, and improving the global exploration efficiency by 3-5 times. The sequential global-local optimization architecture is designed to organically integrate the global exploration ability of Bayesian optimization and the constraint convergence characteristics of local grid search method. Based on the two-stage Bayesian-local grid search hybrid optimization strategy, the global exploration and local convergence are coordinated to solve the convergence contradiction and the mutual exclusion problem of high-dimensional space search efficiency and local extreme escape. Through the Bayesian-local grid search hybrid strategy, the probability agent model guides the global coarse calibration, and then realizes high-precision parameter fine tuning through constrained local optimization, effectively avoiding the convergence instability problem caused by numerical stiffness in traditional gradient method.

[0083] In addition, in the present application, an optimization objective function based on region perception is introduced, and a region perception minimum-maximum optimization objective is embedded to impose hard constraints on the voltage tracking error of safety sensitive intervals such as full charge and deep discharge, to ensure that the lithium extraction risk prediction error is less than 1% under extreme working conditions, and the lithium ion concentration prediction error is less than 1% under full charge state, laying a model foundation for accurate management and control of the safety boundary of energy storage system. Finally, through the deep adaptation of algorithm architecture and engineering requirements, the parameter identification technology of high-capacity lithium ion battery is promoted from "offline coarse calibration" to "online high-precision" paradigm transition. Through the present application, high-precision and cross-condition robust identification of high-capacity lithium ion battery electrochemical parameters is realized, which provides reliable technical support for real-time state estimation and safety management of grid-level energy storage system, and can greatly improve the working condition generalization ability and safety warning reliability of grid-level energy storage battery model, so that the parameter identification efficiency in dynamic scenarios such as energy time shift and frequency modulation response is improved by more than 5 times, providing high-precision bottom model support for real-time state estimation and life prediction of kilowatt-level energy storage power station.

[0084] By adopting the multi-level collaborative optimization mechanism of the application, the parameter identification period across the working condition (0.1C-1C) is shortened from 5-10 hours to 1 hour or less, and through the accuracy enhancement of the safety sensitive interval, the voltage curve form capture accuracy is improved by 65% under the 0.1C-1C fast charging working condition, providing a high robustness model basis for real-time state estimation and life prediction of kilowatt-level energy storage power stations, and at the same time, stimulating the battery design party to optimize the electrode material and geometric configuration to adapt to the sequential identification framework, and improving the intrinsic safety of high-capacity batteries under extreme working conditions from the root.

[0085] Figure 2 A schematic diagram of a DFN electrochemical model is shown. As shown in Figure 2 and Figure 3a , where, inside the solid-phase active material particle, lithium ion diffusion follows Fick's second law, which can be expressed in spherical coordinates as:

[0086]

[0087] where, represents the rate of change of the solid-phase lithium ion concentration with respect to time, t represents time, c s (r,t) is the solid-phase lithium ion concentration, which represents the lithium ion concentration at a position r from the center of the sphere at time t in the spherical coordinate system, D s is the solid-phase diffusion coefficient, which applies the flux boundary condition at the surface of the particle, reflecting the ability of lithium ions to diffuse inside the solid-phase active material particle, r represents the radial coordinate in the spherical coordinate system, which is used to determine the position inside the particle, describes the change of the diffusion flux of lithium ions inside the solid-phase particle with respect to the radial position due to the concentration gradient.

[0088] The lithium ion transport in the electrolyte is described by the concentrated solution theory:

[0089]

[0090] where, ε e represents the volume fraction of the electrolyte, reflecting the volume proportion of the electrolyte in the electrode material, c e (x,t) is the electrolyte lithium ion concentration, which refers to the concentration of lithium ions in the electrolyte at position x in the Cartesian coordinate system at time t, D e,eff is the effective ion diffusion coefficient, x is the axial coordinate in the Cartesian coordinate system, which is used to determine the position in the electrolyte, is the transference number of lithium ions, F is the Faraday constant, j Li is the volume source term of lithium ions.

[0091] The charge conservation in the solid phase and the electrolyte phase follows Ohm's law, which controls the solid-phase potential φs (x, t) and electrolyte potential φ e (x, t) is distributed. The connection between ionic and electronic current is realized through electrochemical reactions at the solid-liquid interface, which is described by the Butler-Volmer equation:

[0092]

[0093] where j Li is the volume source term of lithium ions, representing the production or consumption rate of lithium ions in electrochemical reactions, a s represents the specific surface area, the exchange current density i0depends on the reaction rate constant related to electrolyte concentration, solid phase surface concentration and temperature, a a represents the anode transfer coefficient, a c represents the cathode transfer coefficient, the overpotential η is the driving force of the reaction, R g represents the gas constant, and T represents the temperature.

[0094] For large-capacity lithium-ion batteries, thermal behavior has an important influence on performance, aging and safety. The thermal model is coupled with the electrochemical process through the energy balance equation:

[0095]

[0096] where ρ represents the density of battery materials, C p represents the specific heat capacity, represents the rate of change of temperature with time, λ represents the thermal conductivity, represents the gradient operator, represents the heat conduction term, and the total heat generation Q total includes the reaction heat, the entropy heat, and the ohmic heat in the solid and electrolyte phases.

[0097] The above Fick's second law formula, the formula describing the concentrated solution theory, the Butler-Volmer equation, and the thermal model formula jointly constitute the DFN model. The model input parameters of the DFN model can be divided into static parameters (geometric size, structural characteristics, volume fraction, etc., such as the thermal conductivity can be regarded as an attribute value related to the static parameter) and dynamic parameters (transport characteristics, conductivity, kinetic parameters, etc., such as the total heat generation can be regarded as related to the dynamic parameter such as the reaction heat), most of the dynamic parameters show temperature dependence following the Arrhenius relationship.

[0098] As Figure 3a and Figure 3bTo achieve fine modeling and high-fidelity physical parameter identification of large-capacity lithium-ion batteries, the application proposes a parameter identification method framework that integrates experimental data-driven, global sensitivity analysis, and multi-level optimization strategy. The method is based on the DFN electrochemical model, fully explores the influence mechanism of different parameters on the model output response, and realizes efficient and robust parameter estimation. First, real discharge test data are collected in step S100, then a complete DFN model is built based on the real discharge test data in step S200, and based on this, global sensitivity analysis across conditions is carried out in step S300. The Sobol method is used to quantitatively evaluate the influence of model input parameters on voltage response, and global sensitivity indicators of multiple dimensions such as initial state parameters, transmission parameters, structural parameters and geometric parameters are obtained.

[0099] In the parameter optimization phase, in view of the non-uniformity of the operating state of the large-capacity battery and the local uncertainty of the electrochemical reaction, the application designs a target function based on key region constraints in step S400 to realize differentiated control of the key performance interval. Subsequently, the application performs iterative approximation and optimal parameter search on the target function in step S500 based on the Bayesian optimization method of the Gaussian process, dynamically adjusts the search step length combined with the trust region management strategy, and realizes efficient exploration of the high-dimensional parameter space. To further improve the local search capability and enhance the adaptability to complex (multi-peak, multi-saddle point) target functions, the application introduces a particle size adaptive local search strategy, constructs a parameter perturbation set in stages, and realizes compensation optimization through local fine tuning in the case of failure in the Bayesian optimization phase, thereby ensuring the stability of parameter identification and the global near-optimality of the final solution.

[0100] The following specifically introduces the cross-condition parameter global sensitivity analysis in step S300.

[0101] In view of the multi-parameter coupling and nonlinear interaction effect problems existing in the model parameter identification process of high-capacity lithium-ion batteries, the application proposes a systematic parameter screening method based on weighted Sobol global sensitivity analysis. This method quantifies the influence weight of model parameters on output characteristics under different working conditions by constructing a multi-dimensional response function, significantly improving the efficiency and accuracy of parameter identification of complex electrochemical systems.

[0102] In this embodiment, for the DFN parameter identification problem under multi-rate discharge conditions, a global sensitivity analysis framework based on variance decomposition is established. In step S300, the global sensitivity parameters of the battery electrochemical model are obtained by cross-condition sensitivity analysis, including the following steps:

[0103] The model output corresponding to the battery electrochemical parameters under various discharge rates is calculated, and the multi-objective sensitivity indicators are obtained according to the model output under various discharge rates;

[0104] The global sensitivity analysis method is used to quantitatively evaluate the sensitivity index of each model input parameter of the battery electrochemical parameter to the model output index, wherein the sensitivity index includes the sensitivity index (first-order Sobol sensitivity index S i ) of a single model input parameter to the voltage response and the sensitivity index (total-order sensitivity index ST i ) of the parameter synergy to the voltage response, which respectively represent the independent influence of a single parameter on the output variable (voltage response) and the synergy effect of each parameter with other parameters, and the model output index Y is determined according to the multi-objective sensitivity index;

[0105] The formulae of the first-order Sobol sensitivity index S i and the total-order sensitivity index ST i are as follows:

[0106]

[0107] Wherein Y represents the model output variable, such as the terminal voltage response of the battery, Var(Y) represents the total variance of the output variable, θ i represents the i-th model input parameter, θ ～i represents a vector composed of all other input parameters except θ i , E(Y| θ i ) represents the conditional expectation of the output variable Y when the parameter θ i is fixed, represents the variance of the conditional expectation E(Y| θ i ) about θ i , Var(Y| θ ～i ) represents the conditional variance of the output variable Y when θ ～i is fixed, represents the expectation of the conditional variance Var(Y| θ ～i ) about θ ～i ; S i quantifies the proportion of the output variance caused by the independent action of the parameter θ i , and ST i reflects the total contribution degree including the parameter interaction;

[0108] The multi-dimensional parameter space is constructed through Monte Carlo numerical sampling, and the sensitivity ranking of each parameter to the terminal voltage response under different discharge rate conditions is analyzed, thereby providing a scientific basis for subsequent parameter identification;

[0109] According to the sensitivity index of each model input parameter of the battery electrochemical parameter, the global sensitivity parameters in the model input parameters are screened.

[0110] Traditional sensitivity analysis usually focuses on parameter sensitivity evaluation under a single operating condition, thus can lead to a parameter set that performs well at one specific discharge rate but poorly at other discharge rates. For high-capacity batteries used in grid energy storage applications, it is crucial to maintain performance consistency at different discharge rates. To address this issue, a multi-objective sensitivity index is formulated in this embodiment to replace the battery terminal voltage at a single discharge rate as the model output variable Y when calculating the sensitivity index.

[0111] In this embodiment, the model outputs corresponding to various discharge rates are calculated based on the battery electrochemical parameters, and the multi-objective sensitivity index is obtained based on the model outputs corresponding to various discharge rates, including the following steps:

[0112] The model outputs corresponding to various discharge rates are determined according to the root mean square error between the simulated voltage curves and the experimental voltage curves corresponding to various discharge rates based on the battery electrochemical parameters, and the formula for calculating the model output at the i-th discharge rate C i is as follows:

[0113]

[0114] where N i is the number of time points in the discharge curve at the i-th discharge rate, V sim,j is the voltage value at the j-th time point predicted by the model, and V exp,j is the voltage value at the j-th time point measured experimentally.

[0115] The multi-objective sensitivity index Y multi is calculated based on the model outputs corresponding to various discharge rates and the relative importance weight vector of different discharge rates, and the formula for calculating the multi-objective sensitivity index Y multi is as follows:

[0116]

[0117] where w = [w1, w2, …, w m ] is the weight vector reflecting the relative importance of different discharge rates.

[0118] In summary, based on the parameter sensitivity ranking, combined with the threshold setting of main effect and interaction effect contribution, the dominant parameter set that significantly affects the model output is selected. Parameter identification on this parameter subset can significantly reduce the optimization dimension and improve the global optimization efficiency. For example, by setting the main effect threshold and the interaction effect threshold, when the first-order Sobol sensitivity index of a parameter is greater than the main effect threshold and the total-order sensitivity index is greater than the interaction effect threshold, the parameter is considered as a global sensitivity parameter, and parameters that do not meet the conditions are excluded, thereby realizing parameter dimension reduction.

[0119] The following is a detailed introduction to the establishment of the objective function in step S400 of this embodiment and the implementation of the sequential global-local optimization strategy in step S500. Figure 3b As shown in the figure, after obtaining the main electrochemical parameter set, in order to overcome the dual challenges of large-capacity lithium-ion battery electrochemical model parameter dimensions, large operating condition differences, and concentrated errors in extreme state of charge sections, this application proposes a "sequential global-local optimization strategy". This strategy uses the highly sensitive parameter subset (a set of global sensitivity parameters) obtained from the cross-operating condition global sensitivity analysis as the core identification object, and integrates "global consistency" and "key section priority" into the same optimization framework. The overall process includes four consecutive stages: parameter simplification, regional optimization, trust region constrained Bayesian exploration, and granular adaptive local optimization. It takes into account the global feasibility of the parameter space and ensures the prediction accuracy at the high-low SOC (State of Charge) end.

[0120] In this embodiment, in step S400, constructing an objective function based on key area constraints based on experimental data includes the following steps:

[0121] Based on the discharge rate C across m i Experimental data for (i=1,…,m) The following formula is used to construct the working condition error function based on the root mean square error as the objective function:

[0122]

[0123] in, represents the working condition error, θ represents the global sensitivity parameter vector to be optimized, Indicates the discharge rate C at the i-th i The experimental data set when N i Indicates the discharge rate C i The number of sampling time points of the experimental data is The battery electrochemical model based on the parameter θ is expressed as i Time t j The predicted terminal voltage when Indicates the discharge rate C i Time t j The experimental measurement terminal voltage at ;

[0124] The minimax robust objective is constructed as follows:

[0125]

[0126] Where Θ represents the feasible space of the global sensitivity parameter vector to be optimized;

[0127] The following regional constraint is imposed on the objective function based on the min-max robust objective:

[0128]

[0129] wherein, is a sub-data set of the high state of charge interval in the experimental data set, is a sub-data set of the low state of charge interval in the experimental data set, and ε high and low respectively represent the maximum allowed terminal voltage deviation corresponding to the high state of charge interval and the maximum allowed terminal voltage deviation corresponding to the low state of charge interval, for example, both are set to 10 mV, but the application is not limited thereto. The constraint structure forces the error of each working condition to converge uniformly and imposes an equal accuracy threshold on the SOC end section, avoiding the imbalance phenomenon of "overall error pressure low-end error sacrifice" in the traditional weighted sum method. In this embodiment, the target function based on the regional perception constraint imposes a hard error constraint on the SOC extreme interval, greatly reducing the prediction deviation of the lithium concentration gradient under full charging and deep discharging conditions, and directly supporting the minute-level early warning capability of the battery management system for lithium precipitation risk.

[0130] Therefore, in this embodiment, a target function based on key region constraints is designed, and different error tolerances are introduced for the high state of charge (SOC) and the low state of charge interval, respectively, to realize differentiated control of the key performance interval.

[0131] In this embodiment, in step S500, the trust region-based constrained Bayesian optimization method and the granular adaptive local search method are alternately used to explore the optimization target function, and the optimization target function is iteratively optimized to obtain the optimal solution of the global sensitivity parameter, including performing the following steps in each iteration round:

[0132] The trust region-based constrained Bayesian optimization method is used to globally explore the optimization target function to obtain the global optimal solution of the current iteration round;

[0133] Based on the granular adaptive local search method, the current global optimal solution is adjusted based on the current global optimal solution to obtain a local optimal solution as the basis for constructing the trust region in the next iteration round.

[0134] To efficiently solve the target function within a limited simulation budget, the application uses trust region-constrained Bayesian optimization (TR-CBO) for global exploration. Specifically, the trust region-based constrained Bayesian optimization method is used to globally explore the optimization target function to obtain the global optimal solution of the current iteration round, including the following steps:

[0135] First, define the target function f(θ) on which global exploration is based as follows:

[0136]

[0137] wherein, denotes the maximum value for the index i, i.e. for different discharge rates C i , the error function corresponding to the operating condition is calculated and the maximum value thereof is taken as the objective function f(θ), c H (θ) denotes the constraint function for high SOC, c L (θ) denotes the constraint function for low SOC;

[0138] Gaussian process priors are respectively constructed on the global parameter vector; specifically, Gaussian process priors of zero mean Matérn-5 / 2 kernel are respectively constructed on the parameter vector

[0139]

[0140] wherein, and respectively denote Gaussian processes, the mean functions of which are 0, and the kernel functions of which are k f and f denotes the objective function, j denotes an index symbol, and denotes different constraint conditions, H denotes the constraint of high state of charge region, and L denotes the constraint of low state of charge region;

[0141] Figure 3b The application formula of the Gaussian process prior in this embodiment is shown in the following formula, wherein f(θ) denotes the objective function, m(θ) denotes the prior mean of the objective function, which is usually set as a constant such as 0, k(θ, θ') is a kernel function, which measures the similarity of two points θ and θ' in the parameter space, k f (θ, θ') denotes the kernel function of the objective function, denotes the kernel function of the constraint function, which determines the smoothness and correlation of the Gaussian process;

[0142] In the first round, Latin hypercube design is used for initial sampling to obtain initial sampling points, i.e. n0points are obtained by initial sampling of Latin hypercube design to obtain priori posteriori, and the trust region is determined according to the initial sampling points and the initial trust region radius, and in the subsequent rounds, the trust region is determined based on the trust region radius and the local optimal solution updated in the previous round as follows:

[0143]

[0144] wherein, T n denotes the trust region of the nth iteration, denotes the current optimal parameter vector of the nth iteration, θ denotes the global sensitivity parameter vector to be optimized, denotes the vector for the nth iteration, ​ the infinity norm of the trust region radius L n denotes the trust region radius of the n-th iteration;

[0145] The candidate optimal solution is obtained by minimizing the expected improvement (EI) within the trust region, the global optimal solution is updated by the candidate optimal solution, and the trust region radius is updated;

[0146] Specifically, the expected improvement (EI) formula is as follows:

[0147]

[0148] where θ n+1 is the candidate parameter vector of the n+1-th iteration, arg min denotes finding the parameter θ that minimizes the function value of the EI formula within the trust region, f min denotes the global optimal objective function value found in the current iteration, μ f (θ) and σ f (θ) respectively denote the posterior mean and standard deviation of the objective function at the parameter point θ, [f min -μ f (θ)] + denotes the truncation function, taking the maximum value of [f min -μ f (θ)] + and 0, Z denotes the normalized improvement, Z = [f min -μ f (θ)] / σ f (θ), Φ(Z) and φ(Z) respectively denote the cumulative normal distribution function and the standard normal distribution function, and denote the constraint function c j at the parameter point; if the new point is feasible and improves the global optimal value, the optimal solution is updated and the trust region radius is updated according to the following formula:

[0149]

[0150] where L max denotes the preset maximum trust region radius, τ s and τ f respectively denote the preset success threshold and failure threshold, n s and n f respectively represent the number of successes and failures here, which are used to trigger the expansion or contraction of the trust region radius.

[0151] In this embodiment, the trust region is adaptively expanded or contracted to realize the dynamic migration of the "global-local" focus.

[0152] In this embodiment, when TR-CBO converges to the error step, the algorithm enters the granular adaptive local search (GALS) phase to obtain the local optimal solution. Specifically, in the adaptive search phase, the current global optimal solution is adjusted based on the granular adaptive local search method, with the current global optimal solution as the center, to obtain the local optimal solution, including the following steps:

[0153] Centered on the current global optimal solution, the grid for the search sub-stage is constructed as follows:

[0154]

[0155] Among them, M k represents the search sub-stage grid, θ k represents the current global optimal solution, represents the grid perturbation amplitude of the search sub-stage, z represents an integer vector, represents a p-dimensional integer space, where p is the number of battery electrochemical parameters to be optimized;

[0156] Generate a finite batch of grid candidates in the search sub-phase, and amplify if there are any improvements. And update θ k If the search stagnates, that is, the search fails (IF FAIL), it enters the finalization sub-stage as follows:

[0157]

[0158] Among them, D k For Zhengding Zhangji, is the polarization step size of the final sub-stage, d represents the direction vector of the positive definite tensor concentration, if there is still no improvement, follow Shrinking the grid and polarization step size to achieve "coarse-fine grain" asymptotic approximation, τ is the shrinkage coefficient, which is used to control the decay rate of the grid search step size; after finite iterations, Stop searching and get the local optimal solution θ ＊ .

[0159] Therefore, in the entire sequential process, TR-CBO is responsible for probabilistic global exploration driven by information entropy to quickly locate feasible regions; GALS acts as a deterministic direct searcher to perform differential refinement on the randomly sampled rough solution. The alternating execution of the two can fully utilize every call of the expensive electrochemical model simulation, reducing the total number of iterations from 10 in traditional pure random global optimization to 100. 3 ~10 4 Level down to 10 2 At the same time, the RMSE of the maximum terminal voltage at each rate is limited to within 12mV.

[0160] The implementation process of the battery electrochemical parameter identification method is specifically introduced below with a specific example. In this example, ESM57280AS1 lithium ion battery with rated capacity of 280 Ah is taken as the experimental object. The battery uses lithium iron phosphate positive electrode material and is specially used for large-scale energy storage applications. The battery is prismatic structure, with nominal voltage of 3.2 V, and the modular battery pack provides 16.13 kWh of energy.

[0161] Firstly, corresponding to step S100, the implementation process of obtaining the experimental data obtained by discharging test of the battery at multiple discharge rates includes: performing the experiment under constant temperature condition of 25°C, controlling the voltage window in the range of 2.5-3.35 V, using standardized test sequence, including controlled constant current-constant voltage charging protocol, and then respectively performing discharging test at multiple rates (0.1C, 0.2C, 0.33C and 1C).

[0162] Step S200 is used to construct the battery electrochemical model (DFN model) based on the experimental data, and then step S300 is used for cross-condition global sensitivity analysis. Figure 4 and Figure 5 The cross-condition global sensitivity analysis results are shown, in which the number of winding layers has the highest direct influence (D i = 0.134) and total influence which indicates its key role in determining the capacity of the battery. The negative electrode thickness and the negative electrode active material volume fraction show moderate direct effect but significant total influence, highlighting significant parameter interaction within the DFN model.

[0163] In Figure 4 and Figure 5 , the corresponding parameters from top to bottom are: negative electrode initial lithium ion concentration, positive electrode initial lithium ion concentration, negative electrode diffusion coefficient, positive electrode Bruggeman coefficient, positive electrode diffusion coefficient, positive electrode conductivity, negative electrode conductivity, negative electrode Bruggeman coefficient, negative electrode maximum lithium ion concentration, negative electrode porosity, positive electrode particle radius, positive electrode porosity, separator porosity, negative electrode particle radius, positive electrode maximum lithium ion concentration, positive electrode active material volume fraction, negative electrode active material volume fraction, positive electrode sheet thickness, electrode width, electrode height, negative electrode sheet thickness, and number of electrode sheet winding layers. Based on the sensitivity analysis results, a hierarchical classification of parameters is established: the main parameters include the number of winding layers, the negative electrode thickness, the negative electrode active material volume fraction, etc., which have high sensitivity across discharge rates; the secondary parameters include the positive electrode thickness, the Bruggeman coefficient, etc., which have moderate sensitivity. Then both the main parameters and the secondary parameters are used as global sensitivity parameters for subsequent identification.

[0164] Figure 6 and Figure 7The discharge curves of batteries 81 and 82 under multi-rate discharge conditions are shown. The identified DFN model parameters exhibit excellent performance under 0.1C, 0.2C, 0.33C, and 1C discharge conditions, as demonstrated by the multi-rate discharge validation. Even under the most demanding 1C discharge condition, the model maintains good performance with only a slight increase in error at the end of discharge.

[0165] Figure 8 is a comparison chart of discharge results of different parameters under 0.33C according to an embodiment of the present application, Figure 9 is a comparison chart of discharge results of different parameters under 1C according to an embodiment of the present application. Wherein, EXP represents Experimental Data, SBO represents Standard Bayesian Optimization, SGLO represents Sequential Global-Local Optimization, and RA-SGLO represents Region-Aware Sequential Global-Local Optimization. Figure 8 and Figure 9 shows the effectiveness of the designed method using ablation experiments. As shown in Figure 8 and Figure 9 , the Region-Aware Sequential Global-Local Optimization (RA-SGLO) method exhibits the best performance in all operating regions, with a maximum root mean square error of 12.4 mV, a root mean square error of 7.8 mV in the high SOC region, and a root mean square error of 9.4 mV in the low SOC region, significantly outperforming the Standard Bayesian Optimization (SBO) and Sequential Global-Local Optimization (SGLO) methods.

[0166] Figure 10 shows the result chart of dynamic current condition test. As shown in Figure 10a -d, the dynamic current distribution validation further confirms the robustness of the identified parameters. Figure 10a -d respectively shows the performance of voltage, frequency, current, and SOC. From Figure 10a -d, it can be seen that under the representative dynamic current distribution containing discharge pulses and intermittent charging, the predicted voltage distribution closely tracks the experimental measurement, especially during the rapid current reversal, exhibiting high fidelity.

[0167] In summary, the core breakthrough of the present application is to construct a sequential optimization architecture for cross-condition identification of electrochemical parameters of high-capacity lithium-ion batteries. Through deep coupling of cross-condition sensitivity analysis and regional perception constraints, the mutual exclusion problem of traditional methods in precision, efficiency and generalization is solved. The key technical breakthroughs of the present application are: a parameter dimension reduction mechanism based on dynamic response characteristics of multiple working conditions, which analyzes the core parameter subset that has a dominant effect on the shape of the charge and discharge curve from a high-dimensional mathematical model; a two-stage Bayesian-local grid search hybrid optimization strategy is adopted, which guides efficient exploration of the parameter space through a global probability proxy model, and then realizes higher accuracy calibration through constrained local convergence; and a regional perception constraint objective function design is added, which applies asymmetric weight constraints to the voltage tracking error in safety-sensitive intervals such as full charge and deep discharge, ensuring that the parameter prediction error under different working conditions is less than 1%. The synergistic effect of the above technical chain makes the parameter identification efficiency of grid-level energy storage battery models in 0.1C-1C rate switching scenarios increase by more than 5 times, and the root mean square error (RMSE) of cross-condition voltage prediction is reduced to 1 / 3 of that of traditional methods.

[0168] The multi-level collaborative architecture and its engineering implementation logic of the high-capacity battery parameter identification method proposed in the present application specifically include: a parameter space dimension reduction criterion driven by cross-condition sensitivity analysis, a dynamic screening mechanism based on the strength of multi-physical field coupling and the sensitivity of working condition response; a hybrid solver design of sequential execution of Bayesian optimization and local grid search method, including sparse covariance matrix construction strategy in the global exploration stage, and regularization method for ill-conditioned Hessian matrix in the local convergence stage; and an objective function construction process based on regional perception constraints, involving state of charge interval boundary identification, error weight nonlinear distribution function, and safety threshold linkage correction strategy. Finally, a full-chain technical solution is formed from multi-source working condition data input, parameter space compression, sequential optimization solution to safety boundary control, providing high-precision and strong-robustness bottom model support for real-time state estimation of grid-level energy storage systems.

[0169] As shown in Figure 11 The present application also provides a battery electrochemical parameter identification system for implementing the battery electrochemical parameter identification method of the first aspect, the system comprising:

[0170] A data acquisition module M100 is configured to obtain experimental data obtained by performing discharge tests on the battery at multiple discharge rates;

[0171] A model construction module M200 is configured to construct a battery electrochemical model based on the experimental data;

[0172] The sensitivity analysis module M300 is configured to perform cross-condition sensitivity analysis on the model input parameters of the battery electrochemical model by using a global sensitivity analysis method to obtain global sensitivity parameters.

[0173] The parameter identification module M400 is configured to construct an objective function based on a key region constraint based on experimental data, and alternately use a constraint Bayesian optimization method based on a trust region and a granular adaptive local search method to explore the optimization objective function, iteratively optimize the objective function, and obtain an optimal solution of the global sensitivity parameters.

[0174] In the battery electrochemical parameter identification system of the present application, the functions of each module can be implemented by using the specific embodiments of the battery electrochemical parameter identification method as described above, which will not be repeated here. The battery electrochemical parameter identification system of the present application can obtain the technical effects of the battery electrochemical parameter identification method described above, which will not be repeated here.

[0175] The present application also provides a battery electrochemical parameter identification device, comprising a processor, a memory having executable instructions of the processor stored therein, wherein the processor is configured to execute the steps of the battery electrochemical parameter identification method by executing the executable instructions.

[0176] Those skilled in the art can understand that various aspects of the present application can be implemented as a system, a method or a computer program product. Therefore, various aspects of the present application can be embodied as a complete hardware embodiment, a complete software embodiment (including firmware, microcode, etc.), or an embodiment combining hardware and software aspects, which can be collectively referred to as "circuitry", "module" or "platform".

[0177] The electronic device 600 according to this embodiment of the present application will be described below with reference to Figure 12 Figure 12 The electronic device 600 shown is merely an example and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0178] As shown in Figure 12 The electronic device 600 is in the form of a general computing device. The components of the electronic device 600 can include, but are not limited to, at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different system components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0179] ​The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 performs the steps of various exemplary embodiments of the present application described in the battery electrochemical parameter identification method section above. For example, the processing unit 610 can perform the following steps: Figure 1 Follow the steps shown in .

[0180] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache memory unit 6202 , and may further include a read-only memory unit (ROM) 6203 .

[0181] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include an implementation of a network environment.

[0182] Bus 630 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0183] The electronic device 600 can also communicate with one or more external devices 700 (e.g., a keyboard, a pointing device, a Bluetooth device, etc.), one or more devices that enable a user to interact with the electronic device 600, and / or any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., a router, a modem, etc.). Such communication can occur via an input / output (I / O) interface 650. Furthermore, the electronic device 600 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via a network adapter 660. The network adapter 660 can communicate with other modules of the electronic device 600 via the bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with the electronic device 600, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0184] In the battery electrochemical parameter identification device, the program in the memory implements the steps of the battery electrochemical parameter identification method when the processor executes the program. Therefore, the device can also obtain the technical effects of the battery electrochemical parameter identification method.

[0185] The exemplary embodiments of the present application also provide a computer program product. The computer program product comprises a computer program which, when executed by a processor, implements the steps of the battery electrochemical parameter identification method described above.

[0186] In an embodiment, the computer program product can be a tangible product containing the computer program, such as a computer readable storage medium storing the computer program. The readable storage medium can be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, etc. signals, including but not limited to random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory (Flash), mechanical hard disk (HDD), solid state disk (SSD), etc. Exemplarily, the computer program product can be implemented as a non-volatile storage medium storing the computer program, such as a read-only memory, a Nand Flash, etc.

[0187] In an embodiment, the computer program product can be an intangible product containing the computer program. Exemplarily, the computer program product can be implemented as a virtual digital product, such as an executable file, an installation package, etc. digital file storing the computer program.

[0188] The code of the computer program can be written in one or more programming languages. The programming languages include C language, Java, C++, Python, etc. The program code can be executed entirely on the user computing device, or partially on the user computing device, or as a separate software package, or partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case involving a remote computing device, the remote computing device can be connected to the user computing device through any kind of network, such as a local area network (LAN), a wide area network (WAN), etc., or can be connected to an external computing device (for example, through an Internet connection provided by an operator).

[0189] The computer program can be carried or transmitted by electrical, magnetic, optical, electromagnetic, infrared, etc. signals. The electronic device can convert the signals carrying the computer program into digital signals, and then run the computer program. When the computer program is running on the electronic device, its code is used to make the electronic device execute (more specifically, can make the processor of the electronic device execute) the method steps of various exemplary embodiments of the present application, such as can execute the steps of the battery electrochemical parameter identification method described above.

[0190] The computer program, when executed by a processor, implements the steps of the battery electrochemical parameter identification method described above, therefore, the computer program product can also obtain the technical effects of the battery electrochemical parameter identification method described above.

[0191] The above description is further detailed in combination with specific preferred embodiments of the present application, and cannot be deemed to limit the specific implementation of the present application to these descriptions. For those of ordinary skill in the art to which the present application belongs, several simple deductions or replacements can be made without departing from the concept of the present application, and all of them shall be deemed to fall within the protection scope of the present application.

Claims

1. A battery electrochemical parameter identification method, characterized in that: The steps include: Obtain experimental data obtained by discharging the battery at various discharge rates; constructing a battery electrochemical model based on the experimental data; A global sensitivity analysis method is used to perform cross-operating-condition sensitivity analysis on the model input parameters of the battery electrochemical model to obtain global sensitivity parameters; constructing an objective function based on key area constraints based on the experimental data; The objective function is explored alternately using a constrained Bayesian optimization method based on a trust region and a granular adaptive local search method, and the objective function is iteratively optimized to obtain the optimal solution of the global sensitivity parameter.

2. The battery electrochemical parameter identification method according to claim 1, characterized in that: Performing a cross-operating-condition sensitivity analysis on the battery electrochemical model to obtain a global sensitivity parameter includes the following steps: Calculating the model output of the battery electrochemical parameters corresponding to various discharge rates, and obtaining a multi-target sensitivity index according to the model output at various discharge rates; A global sensitivity analysis method is used to quantitatively evaluate the sensitivity index of each model input parameter of the battery electrochemical parameter to the model output index, wherein the sensitivity index includes the sensitivity index of a single model input parameter to the voltage response and the sensitivity index of the synergistic effect of the parameters to the voltage response, and the model output index is determined based on the multi-objective sensitivity index; According to the sensitivity index of each model input parameter of the battery electrochemical parameter, the global sensitivity parameters in the model input parameters are screened.

3. The battery electrochemical parameter identification method according to claim 2, characterized in that: Calculating the model output corresponding to the battery electrochemical parameters at various discharge rates, and obtaining a multi-target sensitivity index according to the model output at various discharge rates, comprises the following steps: Determining the model output corresponding to various discharge rates according to the root mean square error between the simulated voltage curve and the experimental voltage curve corresponding to the battery electrochemical parameters at various discharge rates; The multi-objective sensitivity index is obtained by considering the model outputs at various discharge rates and the relative importance weight vectors of different discharge rates.

4. The battery electrochemical parameter identification method according to claim 1, characterized in that: Constructing an objective function based on key area constraints based on the experimental data includes the following steps: Based on the experimental data, the following formula is used to construct a working condition error function based on the root mean square error as the objective function: in, represents the working condition error, θ represents the global sensitivity parameter vector to be optimized, Indicates the discharge rate C at the i-th i The experimental data set when N i Indicates the discharge rate C i The number of sampling time points of the experimental data is: The battery electrochemical model based on the parameter θ is expressed as i Time t j The predicted terminal voltage when Indicates the discharge rate C i Time t j The experimental measurement terminal voltage at ; The minimax robust objective is constructed as follows: Where Θ represents the feasible space of the global sensitivity parameter vector to be optimized; The following regional constraints are imposed on the objective function based on the minimax robust objective: in, is the sub-dataset of the high state of charge interval in the experimental data set, represents the sub-dataset of the low state of charge interval in the experimental data set, ε high and ε low They respectively represent the maximum allowable terminal voltage deviation corresponding to the high state of charge interval and the maximum allowable terminal voltage deviation corresponding to the low state of charge interval.

5. The battery electrochemical parameter identification method according to claim 1, characterized in that: Alternately using a trust region-based constrained Bayesian optimization method and a granular adaptive local search method to explore the optimization objective function, iteratively optimize the objective function, and obtain the optimal solution of the global sensitivity parameter, including performing the following steps in each iteration round: A constrained Bayesian optimization method based on a trust region is used to globally explore the optimization objective function to obtain the global optimal solution of the current iteration round; Centered on the current global optimal solution, the current global optimal solution is adjusted based on the granular adaptive local search method to obtain the local optimal solution, which serves as the basis for constructing the trust region in the next iterative round.

6. The battery electrochemical parameter identification method according to claim 5, characterized in that: A constrained Bayesian optimization method based on a trust region is used to globally explore the optimization objective function to obtain the global optimal solution for the current iteration round, including the following steps: Construct Gaussian process priors on the global parameter vector respectively; In the first round, Latin hypercube design is used for initial sampling to obtain the initial sampling points. The trust region is determined based on the initial sampling points and the initial trust region radius. In subsequent rounds, the trust region is determined based on the trust region radius updated in the previous round and the local optimal solution. Minimizing the expected improvement in the trust region is used to obtain a candidate optimal solution, the candidate optimal solution is used to update the global optimal solution, and the trust region radius is updated.

7. The battery electrochemical parameter identification method according to claim 5, characterized in that: The method of adjusting the current global optimal solution based on the granular adaptive local search method to obtain the local optimal solution with the current global optimal solution as the center includes the following steps: Centered on the current global optimal solution, the grid for the search sub-stage is constructed as follows: Among them, M k represents the search sub-stage grid, θ k represents the current global optimal solution, represents the grid perturbation amplitude of the search sub-stage, z represents an integer vector, represents a p-dimensional integer space, where p is the number of battery electrochemical parameters to be optimized; Generate a finite batch of grid candidates in the search sub-phase, and amplify if there are any improvements. And update θ k If the search stagnates, it enters the finalization sub-phase as follows: Among them, D k For Zhengding Zhangji, is the polarization step size of the final sub-stage, d represents the direction vector of the positive definite tensor concentration, if there is still no improvement, follow Shrink the grid and polarization step, iterate until Stop searching until a local optimal solution is obtained.

8. A battery electrochemical parameter identification system, characterized in that: For implementing the battery electrochemical parameter identification method according to any one of claims 1 to 7, the system comprises: A data acquisition module is used to obtain experimental data obtained by performing discharge tests on the battery at various discharge rates; A model building module, used for building a battery electrochemical model based on the experimental data; A sensitivity analysis module, configured to perform a cross-operating-condition sensitivity analysis on the model input parameters of the battery electrochemical model using a global sensitivity analysis method to obtain global sensitivity parameters; A parameter identification module is used to construct an objective function based on key area constraints based on the experimental data; and alternately use a constrained Bayesian optimization method based on a trust region and a granular adaptive local search method to explore the optimization objective function, iteratively optimize the objective function, and obtain the optimal solution of the global sensitivity parameter.

9. A battery electrochemical parameter identification device, characterized in that: include: processor; a memory storing executable instructions for the processor; The processor is configured to execute the steps of the battery electrochemical parameter identification method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the battery electrochemical parameter identification method according to any one of claims 1 to 7 are implemented.

Images