A comprehensive all-solid-state battery mechanism model construction method and system

Abstract

The application discloses a kind of comprehensive all-solid battery mechanism model construction method and system, comprising the following steps: step S1, build the electrochemical mechanism model of the electrode-electrolyte interface buffer layer and the aging mechanism caused by active material loss of all-solid battery, for optimizing battery structure;Step S2, build material database, the material database includes the data obtained by literature review and the parameters updated by electrochemical mechanism model, and key parameters are screened using parameter sensitivity analysis, and then battery performance is predicted using deep operator network;Step S3, design battery complex working condition process, based on model simulation battery performance dataset, and applied to battery management.The application builds a unified platform integrating material database, computational modeling and battery management, providing support for the paradigm shift in all-solid battery development.

Description

Technical Field

[0001] This invention relates to the field of all-solid-state batteries, and in particular to a comprehensive method and system for constructing an all-solid-state battery mechanism model. Background Technology

[0002] All-solid-state batteries, as next-generation energy storage devices with high energy density and high safety, have received widespread attention from academia and industry in recent years. Their core advantage lies in using a solid electrolyte instead of a traditional liquid electrolyte, effectively avoiding safety hazards such as leakage and combustion, and promising higher energy density and longer cycle life. However, all-solid-state batteries still face many challenges in practical applications. Firstly, there are interface problems between the electrodes and the solid electrolyte, including high interfacial impedance, uneven lithium deposition and stripping, and chemical and mechanical degradation during charge and discharge. These interface problems severely affect the actual performance and lifespan of the battery, and existing models cannot accurately reflect these interfacial phenomena. Secondly, the development of all-solid-state batteries still heavily relies on experimental trial and error and experience accumulation, lacking systematic theoretical guidance and intelligent design methods, resulting in a lack of comprehensive mechanistic models and easily accessible material parameter-performance databases. Summary of the Invention

[0003] The purpose of this invention is to provide a comprehensive method and system for constructing a mechanism model of all-solid-state batteries, so as to solve the problems mentioned in the background art.

[0004] To achieve the above-mentioned objectives, one aspect of the present invention provides a comprehensive method for constructing a mechanism model of an all-solid-state battery, comprising the following steps:

[0005] Step S1: Build an electrochemical mechanism model that covers the electrode-electrolyte interface buffer layer and the aging mechanism caused by loss of active materials in all-solid-state batteries, in order to optimize the battery structure.

[0006] Step S2: Build a materials database, which includes data obtained from literature review and parameters updated through electrochemical mechanism models. Use parameter sensitivity analysis to screen key parameters, and then use deep operator networks to predict battery performance.

[0007] Step S3: Design a complex battery operating condition process, simulate battery performance dataset based on the model, and apply it to battery management.

[0008] Furthermore, the electrochemical mechanism model in step S1 is based on the governing equations of each domain of this model, and the expressions of the governing equations of each domain are as follows:

[0009] Expression of the control equation within the electrode:

[0010] ,

[0011] The expression for the electrolyte internal control equation is as follows:

[0012] ,

[0013] ,

[0014] Expression of the governing equation within the buffer layer:

[0015] ,

[0016] ,

[0017] Electrochemical reaction equation expression:

[0018] ,

[0019] ,

[0020] ,

[0021] in, and These represent the electrode lithium concentration and diffusion coefficient, respectively. , and These represent the concentration of lithium ions, the diffusion coefficient, and the diffusion coefficient of n ions in the electrolyte, respectively. , and These represent the concentration of lithium ions, the diffusion coefficient, and the diffusion coefficient of n ions in the buffer layer, respectively. It is the carrier generation term. , and These are the proportion of free lithium ions in the electrolyte at equilibrium, the reaction rate constant, and the initial lithium ion concentration, respectively. , and These represent the fractional reaction rate constant of free lithium ions in the buffer layer at equilibrium and the initial lithium ion concentration, respectively. and These are the total current density, cathode current density, and anode current density, respectively. and These are the rate constants for the cathode and anode reactions, respectively. and These represent the maximum and minimum lithium concentrations in the electrode, respectively. and These are the charge transfer coefficients for the cathode and anode, respectively. It is Faraday's constant. The cross-sectional area of ​​the electrode is... This is the charge transfer overpotential. The gas constant is... For temperature;

[0022] The governing equation for aging due to the loss of battery active materials is:

[0023] ,

[0024] in, The cathode thickness in the first cycle. The cathode thickness is calculated for different number of cycles. This represents the loss rate of active material in the cathode. This represents the number of cycles.

[0025] The overpotential control equation is:

[0026]

[0027]

[0028] In the formula, , and These represent the diffusion migration overpotentials caused by lithium ions flowing in the solid electrolyte and buffer layer, and the insertion and deintercalation overpotentials of lithium ions in the electrode material, respectively.

[0029] Furthermore, the battery structure optimization method includes the following steps:

[0030] Step S101: Select discharge capacity and battery discharge platform as two comparison indicators;

[0031] Step S102: Based on the all-solid-state battery mechanism model, the thicknesses of the battery cathode, electrolyte, and buffer layer are changed, and the battery discharge capacity and discharge platform size are compared to select the optimal battery structure.

[0032] Furthermore, in addition to regular literature review, material parameters are continuously updated and improved through simulation verification. A comprehensive database for solid-state batteries was also developed using Matlab R2023b's App Designer. This software utilizes multiple components such as text labels, buttons, edit boxes, and plotting areas, and corresponding callback functions are written for each component to enable interactive functionality. The platform integrates key physical property data and corresponding discharge curves for cathode, anode, electrolyte, and coating, allowing users to easily view all parameters by simply selecting the material from the drop-down list.

[0033] Furthermore, in step S2, the parameter sensitivity analysis for screening key parameters uses backpropagation neural network (BPNN) and random forest model (RF) as surrogate models. The Sobol global sensitivity analysis method is applied to systematically evaluate the contribution of each parameter to the battery performance output, and all parameters are ranked in order of importance based on the first-order sensitivity index and the total sensitivity index.

[0034] Furthermore, the method of predicting battery performance using deep operator networks includes:

[0035] Step S201: Select the 6 parameters with the highest importance as key parameters and use them as input to predict the battery voltage-capacity curve;

[0036] Step S202: Use three branch networks and one backbone network, each network is implemented as a fully connected neural network;

[0037] In step S203, the backbone network receives the battery capacity, the branch network receives six key parameters, and the output is the voltage distribution under different capacity levels.

[0038] Furthermore, the following methods were used for multi-condition data simulation:

[0039] Step S211: In constant current condition (CC), perform full charge and discharge sequentially at rates of 0.1C, 0.2C, and 0.33C.

[0040] Step S212: HPPC (Hybrid Pulse Condition) sets a 0.33C discharge pulse of 1090s every 10% of SOC from 0% to 100%, followed by a rest period of 10900s.

[0041] Step S213: In the DST (Dynamic Stress Test) condition, discharge at 0.1C for 19 minutes, then let stand for 3 minutes, followed by charging at 0.1C for 10 minutes, then let stand for 3 minutes.

[0042] Step S214: Discharge at 0.2C for 19 minutes, then let stand for 12 minutes until the discharge is complete.

[0043] Furthermore, by combining the electrochemical mechanism with the equivalent circuit model, the parameters of the second-order RC (resistor-capacitance) equivalent circuit model are identified. The steps of the particle swarm optimization algorithm are as follows:

[0044] Step S221: Perform initial settings by randomly generating a set of particles, each particle representing a potential solution to the problem;

[0045] Step S222: Calculate the objective function value for each particle and evaluate its fitness.

[0046] Step S223: If the current position of the particle corresponds to a better fitness, then update the position to the optimal position of the particle.

[0047] Step S224: Select the particle with the best fitness among all particles; its corresponding position is the global optimal solution.

[0048] Step S225: Based on the individual particle and the global optimal position, update the velocity and position of each particle. The velocity update formula includes three key factors: inertial weight, individual acceleration term, and social acceleration term.

[0049] Step S226: Repeat steps S22 to S25 until the termination condition is met and then exit.

[0050] Furthermore, the second-order RC model based on electrochemical principles needs to identify the following model parameters: ohmic internal resistance R0, polarization resistances R1 and R2, and polarization capacitances C1 and C2. R0, R1, and R2 can be obtained through decoupling, while the remaining C1 and C2 are identified using the particle swarm optimization algorithm, which reduces the computational load of the algorithm and improves the efficiency of parameter identification.

[0051] Furthermore, the battery state management, based on the identified model parameters, employs the Singular Value Decomposition-Unscented Kalman Filter (SVD-UKF) algorithm to estimate the battery state of charge (SOC), including the following steps:

[0052] Step S301: Determine the initial value of the state. Initial values ​​of posterior state error covariance The calculation formula is as follows:

[0053] ;

[0054] Step S302, for Time-state quantity The posterior probability distribution is sampled and sigma is calculated using the following formula:

[0055] ,

[0056] Where L is the length of the state vector, and the length of the state vector in this invention is 3. The weight values ​​are calculated as follows:

[0057] ,

[0058] in, α is a secondary scaling factor, which is usually set to 0 in unscented transformation. α is a scaling factor that represents the degree of closeness between the sampling point and the mean, and is usually set to 0.01. β is usually set to 2 for optimal selection under Gaussian distribution.

[0059] Step S303, update the prior state value and system variance prediction values The calculation formula is as follows:

[0060] ,

[0061] in, It is the system noise covariance matrix.

[0062] Step S304, for the state variables at time k The prior probability distribution is sampled using sigma, and the calculation formula is as follows;

[0063] ;

[0064] Step S305, update the observations and observed variance predicted values and The calculation formula is as follows:

[0065] ;

[0066] Step S306: Calculate the Kalman gain The calculation formula is as follows:

[0067] ;

[0068] Step S307, Update the posterior state value and posterior state error covariance The calculation formula is as follows:

[0069] .

[0070] Another aspect of the present invention provides a comprehensive all-solid-state battery mechanism model construction system, including a model construction module, a parameter screening module, and a model application module, wherein:

[0071] The model building module is used to build an electrochemical mechanism model covering the aging mechanism caused by the loss of electrode-electrolyte interface buffer layer and active material in all solid-state batteries, which is used to optimize battery structure.

[0072] The parameter screening module is used to build a materials database, which includes data obtained from literature review and parameters updated through electrochemical mechanism models. Parameter sensitivity analysis is used to screen key parameters, and then deep operator networks are used to predict battery performance.

[0073] The model application module is used to design complex battery operating conditions, simulate battery performance datasets based on models, and apply them to battery management.

[0074] Compared with existing technologies, this system and method have the following advantages:

[0075] 1. The method for interface design and battery management of all-solid-state batteries based on mechanism modeling-database framework proposed in this invention has both high precision and strong versatility. It can be applied to various types of all-solid-state battery systems such as inorganic, composite and polymer batteries, and provides a unified and reliable model platform for battery design and optimization of different material systems.

[0076] 2. This invention systematically identifies key parameters that significantly affect battery performance through parameter sensitivity analysis and the DeepONet algorithm, effectively reducing model complexity and computational burden. While ensuring prediction accuracy, it greatly improves modeling and optimization efficiency, providing a clear direction for parameter control in battery design.

[0077] 3. Based on the model parameters identified in this invention, combined with the singular value decomposition-unscented Kalman filter algorithm, a high-precision and robust estimation of the battery state of charge is achieved, which significantly improves the real-time performance and reliability of the battery management system and has important engineering application value. Attached Figure Description

[0078] Figure 1 The flowchart for the all-solid-state battery mechanism model, database, and performance prediction provided in this embodiment is shown.

[0079] Figure 2 This is a verification diagram of an all-solid-state battery. Among them, Figure 2 (a) is a schematic diagram of an all-solid-state battery model including a buffer layer at the battery electrode-electrolyte interface. Figure 2 (b) shows the model verification results of the all-solid-state battery.

[0080] Figure 3 This is a MATLAB app that provides material parameters for all-solid-state batteries, including 2 types of negative electrode materials, 16 types of electrolyte materials, 3 types of buffer layer materials, and 3 types of positive electrode materials, as well as the discharge curves of the combined materials.

[0081] Figure 4 This is a schematic diagram of model parameter sensitivity analysis. Among them, 4(a) is the overall flowchart from surrogate model to parameter analysis, and 4(b) is the DeepONet algorithm structure diagram.

[0082] Figure 5 Figures show the application results of battery models and datasets in battery structure design. Among them, 5(a) is the discharge curve with different battery buffer layer thicknesses, 5(b) is the discharge curve with different battery cathode thicknesses, 5(c) is the discharge curve with different battery electrolyte thicknesses, and 5(d) is a comparison of discharge curve results with different buffer layer thicknesses.

[0083] Figure 6 This document presents a flowchart of the complex charge-discharge cycle for all-solid-state batteries and a schematic diagram of the framework for estimating SOC results using battery models and datasets. 6(a) is the flowchart for the HPPC cycle, 6(b) is the flowchart for the DST cycle, 6(c) is the model diagram for the electrochemical decoupling identification parameter mechanism, 6(d) is the diagram for the second-order RC model, and 6(e) is the diagram for the Kalman filter algorithm. Detailed Implementation

[0084] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0085] The all-solid-state battery mechanism model established in this invention provides a comprehensive and systematic solution for battery structure optimization and battery management. This model comprehensively considers lithium-ion diffusion and migration in the electrodes and electrolytes, interfacial electrochemical reactions, and aging mechanisms caused by active material loss, exhibiting good mechanistic integrity and predictive accuracy. Based on this, this invention introduces parameter sensitivity analysis, using the Sobol global sensitivity analysis method combined with a machine learning surrogate model to systematically identify key parameters that significantly affect battery performance, providing clear direction for parameter control and experimental design basis for performance optimization. Simultaneously, this study integrates the DeepONet architecture, using key parameters as input and voltage-capacity curves as output, achieving high-precision and rapid prediction of battery performance. While maintaining accuracy similar to traditional high-fidelity models, it significantly improves computational efficiency, effectively supporting performance evaluation and material selection under multiple parameter combinations. Furthermore, based on this model, battery performance datasets under multiple operating conditions (CC, HPPC, DST) were generated. By combining the particle swarm optimization algorithm and the SVD-UKF filtering algorithm, high-precision estimation of model parameters and state of charge was achieved, fully demonstrating the application potential of this integrated modeling framework in battery design, state estimation and management strategies.

[0086] like Figure 1 The diagram shown is a flowchart of the all-solid-state battery mechanism model, database, and performance prediction process provided in this embodiment, including the following steps:

[0087] Step S1: Build an electrochemical mechanism model that covers the electrode-electrolyte interface buffer layer and the aging mechanism caused by the loss of active materials in all-solid-state batteries, in order to optimize the battery structure.

[0088] Step S2 involves building a materials database that integrates data obtained from literature reviews and parameters updated through electrochemical mechanism models. Key parameters are then screened using parameter sensitivity analysis, and deep operator networks are used to predict battery performance.

[0089] Step S3 involves designing a complex battery operating condition process, simulating battery performance datasets based on a model, and applying this simulation to battery management. This invention constructs a unified platform integrating a materials database, computational modeling, and battery management, providing support for a paradigm shift in the development of all-solid-state batteries.

[0090] This embodiment constructs a closed-loop framework of "model-database-performance" for all-solid-state batteries. Although the model has been validated in lithium-ion, sodium-ion, and lithium-oxygen batteries, its universality needs to be proven for all-solid-state lithium batteries, and the significant uncertainty of material parameters is the primary obstacle. Therefore, a material database is generated through systematic literature review and continuous coefficient calibration combined with model inversion. In this process, global sensitivity analysis is used to lock key parameters, providing direction for parameter adjustments in experiments and simulations. To further reduce computational costs, a deep operator network, DeepONet, is introduced to predict voltage-capacity curves. Based on this framework, the optimization of battery geometry and high-precision estimation of SOC can be achieved simultaneously.

[0091] like Figure 2 As shown, this embodiment demonstrates model validation. To verify the accuracy and universality of the model, we selected 18 different batteries and simulated the battery discharge process using a constructed all-solid-state battery model. Figure 2 (a) shows the simplified geometry of the model. In this embodiment, the governing equations are assumed to be based on a dense battery structure, with electrochemical reactions occurring only at the electrode-electrolyte interface, which can therefore be considered the current collector. The governing equations within the electrode primarily describe the diffusion behavior of lithium ions, which can be described by Fick's second law. The corresponding concentration cLis is shown below:

[0092] ,

[0093] The corresponding initial values ​​and boundary values ​​are respectively

[0094] ,

[0095] ,

[0096] here and These represent the electrode lithium concentration and diffusion coefficient, respectively.

[0097] In solid electrolytes, lithium ion diffusion and migration are the primary processes. It's important to note that the lithium ions in this process are divided into stationary and mobile ions. This can be represented by the Nernst-Planck equation, which includes charge carriers:

[0098] ,

[0099] ,

[0100] The corresponding initial values ​​and boundary values ​​are respectively

[0101] ,

[0102] ,

[0103] In the formula, , and These represent the concentration of lithium ions, the diffusion coefficient, and the diffusion coefficient of n ions in the electrolyte, respectively. It is the carrier generation term. , and These are the proportion of free lithium ions in the electrolyte at equilibrium, the reaction rate constant, and the initial lithium ion concentration, respectively.

[0104] The buffer layer is essentially an ionic conductor, therefore the diffusion and migration of lithium ions occur within it. It's important to note that these lithium ions are divided into stationary and mobile ions. This can be represented by the Nernst-Planck equation, which includes charge carriers:

[0105] ,

[0106] ,

[0107] The corresponding initial values ​​and boundary values ​​are respectively

[0108] ,

[0109] ,

[0110] In the formula, , and These represent the concentration of lithium ions, the diffusion coefficient, and the diffusion coefficient of n ions in the buffer layer, respectively. , and These represent the fractional reaction rate constant of free lithium ions in the buffer layer under equilibrium conditions and the initial lithium ion concentration, respectively.

[0111] Since the electrochemical reactions in all-solid-state batteries only occur at the electrode-electrolyte interface, if a buffer layer is present, the reactions will occur at the electrode-buffer layer interface, which can be described using the Butler–Volmer equation:

[0112] ,

[0113] ,

[0114] ,

[0115] In the formula, I, and These are the total current density, cathode current density, and anode current density, respectively. and These are the rate constants for the cathode and anode reactions, respectively. and These represent the maximum and minimum lithium concentrations in the electrode, respectively. and These are the charge transfer coefficients for the cathode and anode, respectively. It is Faraday's constant. The cross-sectional area of ​​the electrode is... This is the charge transfer overpotential. The gas constant is... For temperature.

[0116] The model considers the aging phenomenon caused by the loss of active material. Since the negative electrode is lithium metal, it can be treated as a single point. The loss of active material mainly occurs in the positive electrode material. Assuming that the electrode has a dense structure, the loss of active material can be represented by the decrease in electrode thickness with cycling.

[0117] ,

[0118] In the formula, The cathode thickness in the first cycle. The cathode thickness is calculated for different number of cycles. This represents the loss rate of active material in the cathode. This represents the number of cycles.

[0119] The total potential of a battery consists of both overpotential and equilibrium potential. The overpotential in the model includes:

[0120] ,

[0121] ,

[0122] In the formula, , and These represent the diffusion migration overpotentials caused by lithium ions flowing in the solid electrolyte and buffer layer, and the insertion and deintercalation overpotentials of lithium ions in the electrode material, respectively.

[0123] Figure 2 (b) also shows the error between the simulation and experiment of all-solid-state batteries, indicating that the model has good accuracy and universality.

[0124] like Figure 3 As shown, this embodiment constructs a material parameter database for all-solid-state batteries. The database of solid-state battery material and model parameters was developed using the "App Designer" function of Matlab R2023b. The database interface is meticulously designed for intuitiveness and user-friendliness, integrating basic functions such as text labels, buttons, editing fields, and plotting capabilities, and callback functions supporting these functions were written. These functions allow users to easily browse the database, input specific parameters, and visualize data. The app developed in this invention is a comprehensive resource containing parameter values ​​for solid-state battery cathodes, anodes, electrolytes, coatings, etc. It is worth noting that the same material may exhibit inconsistent properties in different batteries. This variability is mainly due to the influence of various factors during battery manufacturing, such as the environment, specific technologies employed, and the instruments used by researchers. To account for these variations, some parameters are set to ranges rather than fixed values, thereby providing a more realistic and flexible representation of material properties.

[0125] To better understand the impact of parameters on battery performance, this embodiment introduces parameter sensitivity analysis. For example... Figure 4 As shown in (a), since the data to be analyzed in this embodiment does not have a clear objective function, a surrogate model is needed to assist in the analysis. First, three key characteristics of discharge performance were selected as the indicators to be analyzed, including total discharge time, which is closely related to the battery capacity; initial discharge voltage, which is related to the battery's SOC state; and the voltage value when the discharge time is half of the total time, which can indirectly represent the battery's discharge plateau. In this invention, a BP neural network and a random forest model are used as surrogate models. Subsequently, the Sobol global sensitivity analysis method is used to analyze the results of the random forest model, using the first-order sensitivity index and the total sensitivity index as indicators for ranking the importance of parameters.

[0126] Furthermore, to improve efficiency while maintaining the accuracy of the all-solid-state battery model, this embodiment extracts six key parameters from 20 parameters based on parameter sensitivity analysis. These parameters include geometric parameters and material property parameters. Subsequently, the embodiment uses DeepONet to predict the battery voltage, such as... Figure 4As shown in (b), three branch networks and one backbone network were used, each implemented as a fully connected neural network. The backbone network receives the battery capacity, and the branch networks receive system parameters, namely six key parameters, and output the voltage distribution at different capacity levels.

[0127] This embodiment can optimize the battery structure and improve battery performance using an all-solid-state battery model. For example... Figure 5 As shown, we change the thickness of the positive electrode, electrolyte, and buffer layer to simulate the discharge curves of batteries with different structures. Using the final discharge capacity and discharge plateau of the battery as indicators, we set the optimal battery structure. Figure 5 (d) The strategy of changing the buffer layer showed extremely high cost-effectiveness. The same discharge performance could be achieved by increasing its thickness by only 1 / 10 and 1 / 100 of that of the electrolyte and positive electrode, respectively.

[0128] Furthermore, to address the issues of poor data sample accessibility and inconsistent formats, this embodiment designs processes for two complex operating conditions, such as... Figure 6 As shown in (ab). First, in CC, full charge and discharge are performed sequentially at 0.1C, 0.2C, and 0.33C rates; HPPC sets a 1090s 0.33C discharge pulse every 10% of SOC from 0% to 100%, followed by a 10900s rest period; in DST condition, discharge at 0.1C for 19 minutes and rest for 3 minutes, then charge at 0.1C for 10 minutes and rest for 3 minutes, and finally discharge at 0.2C for 19 minutes and rest for 12 minutes, until discharge is complete.

[0129] Based on the data sample, this embodiment can estimate the battery's SOC. For example... Figure 6 As shown in (c), the battery SOC is estimated using the SVD-UKF algorithm after identifying parameters through electrochemical decoupling. First, the parameters of the second-order RC equivalent circuit model can be identified using the particle swarm optimization algorithm. The detailed steps are as follows:

[0130] Step S221, Initial setup: Randomly generate a set of particles, each particle representing a potential solution to the problem;

[0131] Step S222: Evaluate fitness: Calculate the objective function value, i.e., fitness, for each particle;

[0132] S223. Update the optimal position of the individual: If the current position of the particle corresponds to a better fitness, then update the position to the optimal position of the particle.

[0133] S224. Update the global optimal position: Select the particle with the best fitness among all particles, and its corresponding position is the global optimal solution.

[0134] S225. Update Velocity and Position: Based on the individual particle and the global optimal position, update the velocity and position of each particle. The velocity update formula includes three key factors: inertia weight, individual acceleration term, and social acceleration term.

[0135] S226. Iterative execution: Repeat steps S22 to S25 until the termination condition is met (such as reaching the preset maximum number of iterations or when the objective function converges to the given accuracy range) and then exit.

[0136] Secondly, by combining the electrochemical mechanism with the equivalent circuit model, the parameter identification process of the model is optimized, making the identification results more consistent with the actual characteristics of the battery. The second-order RC model based on electrochemical principles needs to identify the model parameters of ohmic internal resistance R0, polarization resistances R1 and R2, and polarization capacitances C1 and C2. R0, R1, and R2 can be obtained through decoupling, while the remaining C1 and C2 are identified using the particle swarm optimization algorithm, which reduces the computational load of the algorithm and improves the parameter identification efficiency.

[0137] Finally, the singular value decomposition-unscented Kalman filter algorithm is used to estimate the battery state of charge. The specific steps are as follows:

[0138] Step S301: Determine the initial value of the state. Initial values ​​of posterior state error covariance The calculation formula is as follows:

[0139] ;

[0140] Step S302, for the state variables at time k-1 The posterior probability distribution is sampled and sigma is calculated using the following formula:

[0141] ,

[0142] Where L is the length of the state vector, and the length of the state vector in this invention is 3. The weight values ​​are calculated as follows:

[0143] ,

[0144] in, α is a secondary scaling factor, which is usually set to 0 in unscented transformation. α is a scaling factor that represents the degree of closeness between the sampling point and the mean, and is usually set to 0.01. β is usually set to 2 for optimal selection under Gaussian distribution.

[0145] Step S303, update the prior state value and system variance prediction values The calculation formula is as follows:

[0146] ,

[0147] in, It is the system noise covariance matrix.

[0148] Step S304, for the state variables at time k The prior probability distribution is sampled using sigma, and the calculation formula is as follows;

[0149] ;

[0150] Step S305, update the observations and observed variance predicted values and The calculation formula is as follows:

[0151] ;

[0152] Step S306: Calculate the Kalman gain The calculation formula is as follows:

[0153] ;

[0154] Step S307, Update the posterior state value and posterior state error covariance The calculation formula is as follows:

[0155] .

[0156] Another embodiment of the present invention provides a comprehensive all-solid-state battery mechanism model construction system, including a model construction module, a parameter selection module, and a model application module, wherein:

[0157] The model building module is used to build an electrochemical mechanism model covering the aging mechanism caused by the loss of electrode-electrolyte interface buffer layer and active material in all solid-state batteries, which is used to optimize battery structure.

[0158] The parameter screening module is used to build a materials database, which includes data obtained from literature review and parameters updated through electrochemical mechanism models. Parameter sensitivity analysis is used to screen key parameters, and then deep operator networks are used to predict battery performance.

[0159] The model application module is used to design complex battery operating conditions, simulate battery performance datasets based on models, and apply them to battery management.

[0160] In summary, this invention proposes a method for interface design and battery management of all-solid-state batteries based on a mechanism modeling-database framework. By constructing a coupled model covering ion transport, electrochemical reactions, and aging mechanisms in the electrodes, electrolyte, and buffer layer, a systematic material parameter database is established. Furthermore, a technical approach combining parameter sensitivity analysis and DeepONet is introduced, enabling effective screening of key battery performance parameters and high-precision, high-efficiency voltage curve prediction. This method demonstrates good applicability and robustness under various battery systems and operating conditions, providing clear guidance for all-solid-state battery structure optimization and a reliable data foundation and algorithmic support for its state estimation and system management. It possesses significant theoretical and engineering application value.

[0161] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A comprehensive method for constructing a mechanism model of all-solid-state batteries, characterized in that, Includes the following steps: Step S1: Build an electrochemical mechanism model that covers the electrode-electrolyte interface buffer layer and the aging mechanism caused by loss of active materials in all-solid-state batteries, in order to optimize the battery structure. Step S2: Build a materials database, which includes data obtained from literature review and parameters updated through electrochemical mechanism models. Use parameter sensitivity analysis to screen key parameters, and then use deep operator networks to predict battery performance. Step S3: Design a complex battery operating condition process, simulate battery performance dataset based on the model, and apply it to battery management.

2. The method for constructing a comprehensive all-solid-state battery mechanism model according to claim 1, characterized in that, The electrochemical mechanism model in step S1 is based on the governing equations of each domain of the model. The expressions of the governing equations of each domain are as follows: Expression of the control equation within the electrode: ， The expression for the electrolyte internal control equation is as follows: ， ， Expression of the governing equation within the buffer layer: ， ， Electrochemical reaction equation expression: ， ， ， in, and These represent the electrode lithium concentration and diffusion coefficient, respectively. , and These represent the concentration of lithium ions, the diffusion coefficient, and the diffusion coefficient of n ions in the electrolyte, respectively. , and These represent the concentration of lithium ions, the diffusion coefficient, and the diffusion coefficient of n ions in the buffer layer, respectively. It is the carrier generation term. , and These are the proportion of free lithium ions in the electrolyte at equilibrium, the reaction rate constant, and the initial lithium ion concentration, respectively. , and These represent the fractional reaction rate constant of free lithium ions in the buffer layer at equilibrium and the initial lithium ion concentration, respectively. and These are the total current density, cathode current density, and anode current density, respectively. and These are the rate constants for the cathode and anode reactions, respectively. and These represent the maximum and minimum lithium concentrations in the electrode, respectively. and These are the charge transfer coefficients for the cathode and anode, respectively. It is Faraday's constant. The cross-sectional area of ​​the electrode is... This is the charge transfer overpotential. The gas constant is... For temperature; The governing equation for aging due to the loss of battery active materials is: ， in, The cathode thickness in the first cycle. The cathode thickness is calculated for different number of cycles. This represents the loss rate of active material in the cathode. This represents the number of cycles.

3. The method for constructing a comprehensive all-solid-state battery mechanism model according to claim 1, characterized in that, The battery structure optimization method includes the following steps: Step S101: Select discharge capacity and battery discharge platform as two comparison indicators; Step S102: Based on the all-solid-state battery mechanism model, the thicknesses of the battery cathode, electrolyte, and buffer layer are changed, and the battery discharge capacity and discharge platform size are compared to select the optimal battery structure.

4. The method for constructing a comprehensive all-solid-state battery mechanism model according to claim 1, characterized in that, In step S2, the parameter sensitivity analysis for screening key parameters uses backpropagation neural network and random forest model as surrogate models, and applies Sobol global sensitivity analysis method to rank the importance of all parameters.

5. The method for constructing a comprehensive all-solid-state battery mechanism model according to claim 1, characterized in that, The method of predicting battery performance using deep operator networks includes: Step S201: Select the 6 parameters with the highest importance as key parameters and use them as input to predict the battery voltage-capacity curve; Step S202: Use three branch networks and one backbone network, each network is implemented as a fully connected neural network; In step S203, the backbone network receives the battery capacity, the branch network receives six key parameters, and the output is the voltage distribution under different capacity levels.

6. The method for constructing a comprehensive all-solid-state battery mechanism model according to claim 1, characterized in that, The following methods were used for multi-condition data simulation: Step S211: In constant current condition (CC), perform full charge and discharge sequentially at rates of 0.1C, 0.2C, and 0.33C. In step S212, HPPC sets a 0.33C discharge pulse of 1090s every 10% of SOC between 0% and 100%, followed by a rest period of 10900s. Step S213: Discharge at 0.1C for 19 minutes under DST conditions, then let stand for 3 minutes, followed by charging at 0.1C for 10 minutes, then let stand for 3 minutes. Step S214: Discharge at 0.2C for 19 minutes, then let stand for 12 minutes until the discharge is complete.

7. The method for constructing a comprehensive all-solid-state battery mechanism model according to claim 1, characterized in that, By combining electrochemical mechanisms with equivalent circuit models, the parameters of the second-order RC equivalent circuit model are identified. The steps of the particle swarm optimization algorithm are as follows: Step S221: Perform initial settings by randomly generating a set of particles, each particle representing a potential solution to the problem; Step S222: Calculate the objective function value for each particle and evaluate its fitness. Step S223: If the current position of the particle corresponds to a better fitness, then update the position to the optimal position of the particle. Step S224: Select the particle with the best fitness among all particles; its corresponding position is the global optimal solution. Step S225: Based on the individual particle and the global optimal position, update the velocity and position of each particle. The velocity update formula includes three key factors: inertial weight, individual acceleration term, and social acceleration term. Step S226: Repeat steps S22 to S25 until the termination condition is met and then exit.

8. The method for constructing a comprehensive all-solid-state battery mechanism model according to claim 1, characterized in that, The second-order RC model based on electrochemical principles needs to identify the following model parameters: ohmic internal resistance R0, polarization resistances R1 and R2, and polarization capacitances C1 and C2. R0, R1, and R2 can be obtained by decoupling, while the remaining C1 and C2 are identified by particle swarm optimization, which reduces the computational load of the algorithm and improves the efficiency of parameter identification.

9. The method for constructing a comprehensive all-solid-state battery mechanism model according to claim 1, characterized in that, The battery state management is based on the identified model parameters and uses a singular value decomposition-unscented Kalman filter algorithm to estimate the battery state of charge, including the following steps: Step S301: Determine the initial value of the state. Initial values ​​of posterior state error covariance ; Step S302, for Time-state quantity The posterior probability distribution is used for sigma sampling calculation; Step S303, update the prior state value and system variance prediction values ; Step S304, for the state variables at time k Sigma sampling is performed on the prior probability distribution; Step S305, update the observations and observed variance predicted values and ; Step S306: Calculate the Kalman gain ; Step S307, Update the posterior state value and posterior state error covariance .

10. A comprehensive system for constructing a mechanism model of all-solid-state batteries, characterized in that, It includes a model building module, a parameter filtering module, and a model application module, among which: The model building module is used to build an electrochemical mechanism model covering the aging mechanism caused by the loss of electrode-electrolyte interface buffer layer and active material in all solid-state batteries, which is used to optimize battery structure. The parameter screening module is used to build a materials database, which includes data obtained from literature review and parameters updated through electrochemical mechanism models. Parameter sensitivity analysis is used to screen key parameters, and then deep operator networks are used to predict battery performance. The model application module is used to design complex battery operating conditions, simulate battery performance datasets based on models, and apply them to battery management.

Images