Battery aging prediction method, device, storage medium, and computer program product

Abstract

The embodiment of the application provides a battery aging prediction method and device, a storage medium and a computer program product, and relates to the technical field of batteries. The battery aging prediction method comprises: obtaining first simulation data by using an electrochemical model, and constructing a first prediction model; obtaining a first aging prediction result by using the first prediction model according to the first simulation data; obtaining first test data and constructing a second prediction model; fusing the first prediction model and the second prediction model to form a fusion prediction model; and obtaining a second aging prediction result by using the fusion prediction model according to the first simulation data and the first test data. The application can enable the prediction model to learn the aging law of the battery, improve the prediction accuracy of the battery aging, and avoid the problems of large calculation amount of simulation and possible deviation of test data by combining the simulation data and the test data and using the fusion prediction model for aging prediction, thereby reducing the prediction cost and improving the prediction efficiency.

Description

Technical Field

[0001] This application relates to the field of battery technology, and in particular to a battery aging prediction method, apparatus, electronic device, storage medium, and computer program product. Background Technology

[0002] With the rapid development of battery technology, lithium batteries and other batteries are increasingly widely used in energy storage systems, electric vehicles, and other fields due to their advantages such as high energy density and low cost. Battery aging affects battery performance, safety, and lifespan. Currently, battery aging prediction typically employs data-driven methods based on test data. These methods collect test data such as capacity, voltage, and internal resistance during charge-discharge cycle testing to obtain predictions of battery aging. However, the accuracy of existing data-driven methods depends on the test data; however, data collection is time-consuming and costly, and the test data is subject to limitations imposed by testing conditions, making it prone to bias. Therefore, existing data-driven methods suffer from low accuracy in predicting battery aging. Summary of the Invention

[0003] In view of the above problems, this application provides a battery aging prediction method, apparatus, electronic device, storage medium, and computer program product to improve the accuracy of battery aging prediction.

[0004] According to a first aspect of this disclosure, a battery aging prediction method is provided, characterized by comprising: obtaining first simulation data corresponding to the aging of the battery using an electrochemical model of the battery based on the battery's charge-discharge process and operating parameters corresponding to the charge-discharge process; constructing a first prediction model based on the first simulation data; and obtaining a first aging prediction result of the battery using the trained first prediction model based on the first simulation data.

[0005] In this embodiment, by utilizing the battery's charging and discharging process and operating parameters, as well as the characteristics of the electrochemical model, simulation data can be generated. This allows for accurate simulation of the battery's aging model and the acquisition of precise aging simulation data. The prediction model can then learn the battery's aging patterns, improving the accuracy of battery aging predictions and enhancing the rationality, reliability, and versatility of battery aging predictions. Furthermore, it avoids the need for battery aging tests, reducing prediction costs and improving prediction efficiency.

[0006] In some embodiments, the battery is subjected to an aging test to obtain first test data; a second prediction model is constructed based on the first test data; the first prediction model and the second prediction model are fused to form a fused prediction model; and a second aging prediction result of the battery is obtained using the trained fused prediction model based on the first simulation data and the first test data.

[0007] In this embodiment, a second prediction model is constructed based on the first test data obtained from the battery aging test. A fusion prediction model is generated based on the first and second prediction models. The second aging prediction result of the battery is obtained through the trained fusion prediction model. By combining simulation data and test data and using the fusion prediction model for aging prediction, the large computational load of simulating battery aging and the potential bias of the test data obtained from battery aging tests can be avoided. This avoids the limitations of a single model, improves the prediction accuracy of battery aging, reduces prediction costs, and improves prediction efficiency.

[0008] In some embodiments, before obtaining a second aging prediction result of the battery using the trained fusion prediction model, the method includes: constructing sample data for training the fusion prediction model; training the fusion prediction model using the sample data; and adjusting the parameters of the fusion prediction model according to the fusion loss function value of the fusion prediction model to obtain the trained fusion prediction model.

[0009] In this embodiment, by constructing sample data and training the fusion prediction model based on the fusion loss function, the fusion prediction model can fully learn the effective information from the simulation data and test data, so that the trained fusion prediction model has better generalization ability and can improve the prediction accuracy of battery aging.

[0010] In some embodiments, a prediction average is determined and a corresponding sample label value is obtained based on the sample prediction values ​​of the first prediction model and the sample prediction values ​​of the second prediction model; the absolute value of the prediction difference is determined based on the prediction average and the sample label value; the sum of the absolute value of the prediction difference and the prediction penalty value is used as the fusion loss function value, or the absolute value of the prediction difference is used as the fusion loss function value.

[0011] In this embodiment, the absolute value of the difference can be used to determine the deviation between the predicted value and the true value, thereby improving the sensitivity of the fusion loss function to the prediction error. The prediction penalty value can avoid extreme prediction deviations, enabling the fusion prediction model to converge towards a better direction and improving the prediction accuracy and stability of battery aging.

[0012] In some embodiments, the predicted penalty value includes: a first penalty value; the method includes: determining a first absolute value of difference based on the sample predicted value of the first prediction model and the sample predicted value of the second prediction model; and weighting the first absolute value of difference using a first weighting parameter to obtain the first penalty value.

[0013] In this embodiment, the penalty value can be determined based on the prediction difference between the first and second prediction models and the weighting parameters, which can make the predictions of the first and second prediction models more consistent, reduce prediction conflicts between models, strengthen the coordination between models, and improve the prediction accuracy and prediction stability of battery aging.

[0014] In some embodiments, the predicted penalty value includes a second penalty value; the method includes: determining a second absolute difference value based on the sample predicted value of the first prediction model and the sample label value; determining a third absolute difference value based on the sample predicted value of the second prediction model and the sample label value; and weighting the sum of the second absolute difference value and the third absolute difference value using a second weighting parameter to obtain the second penalty value.

[0015] In this embodiment, the penalty value can be determined based on the absolute value of the difference between the sample predicted value and the sample label value of the first and second prediction models and the weighting parameters. This can simultaneously constrain the deviation between the first and second prediction models and the true value, avoid a single model dominating the prediction result, and enable the first and second prediction models to have independent prediction accuracy, thereby improving the prediction accuracy and prediction stability of battery aging.

[0016] In some embodiments, the sample prediction values ​​of the first prediction model are weighted by a third weighting parameter to obtain a first prediction weighted value; the sample prediction values ​​of the second prediction model are weighted by a fourth weighting parameter to obtain a second prediction weighted value; wherein the sum of the third weighting parameter and the fourth weighting parameter is 1; a prediction weighted sum is determined based on the first prediction weighted value and the second prediction weighted value, and the corresponding sample label value is obtained; the absolute value of the prediction weighted difference is determined based on the prediction weighted sum and the sample label value, and is used as the fusion loss function value.

[0017] In this embodiment, the predicted values ​​of the first and second prediction models are weighted by weighting parameters, and the absolute value of the difference between the sum of the weighted results and the sample label values ​​is calculated as the fusion loss function. This allows for dynamic adjustment of the contribution of the first and second prediction models to the fusion prediction, enabling the fusion prediction model to adaptively optimize the weights, improving the guidance of the fusion loss function for training the fusion prediction model, and enhancing the prediction accuracy of battery aging.

[0018] In some embodiments, the first model sample data includes a first sample and a first label value corresponding to the first sample, and the second model sample data includes a second sample and a second label value corresponding to the second sample; the step of constructing sample data for training the fusion prediction model includes: obtaining second simulation data and aging parameters corresponding to the aging of the battery using the electrochemical model based on the charge / discharge process and the operating condition parameters; constructing the first model sample data based on the second simulation data and the aging parameters, wherein the first sample is the second simulation data and the first label value is the aging parameter; performing the aging test on the battery to obtain second test data and the aging parameters; constructing the second model sample data based on the second test data and the aging parameters, wherein the second sample is the second test data and the second label value is the aging parameter; and constructing the sample data based on the first model sample data and the second model sample data.

[0019] In this embodiment, sample data is constructed based on the second simulation data generated by the electrochemical model, the second test data obtained by the aging test, and the corresponding aging parameters. This sample data includes simulation data that reflects the theoretical laws of battery aging and test data that reflects the actual aging characteristics of the battery. This provides high-quality and comprehensive samples for training the fusion prediction model and improves the prediction accuracy of the fusion prediction model for battery aging.

[0020] In some embodiments, performing the aging test on the battery to obtain second test data and the aging parameters includes: performing an aging test on the battery to obtain a first state of health (SOH) curve of the battery, wherein the first SOH curve characterizes the relationship between the storage time of the battery and SOH; determining the aging parameters in the first SOH curve, and determining the second test data corresponding to the aging parameters, wherein the aging parameters include: the remaining time until the end of energy period (EOL).

[0021] In this embodiment, by using the SOH curve obtained from the battery aging test, the second test data corresponding to the remaining time to the EOL point can be determined, which can obtain accurate second test data and improve the prediction accuracy of the second prediction model for the remaining time to the EOL point.

[0022] In some embodiments, obtaining second simulation data and aging parameters corresponding to the aging of the battery using the electrochemical model based on the charge / discharge process and the operating condition parameters includes: simulating the aging process of the battery using simulation tools and the electrochemical model based on the charge / discharge process and the operating condition parameters to obtain a second state of health (SOH) curve of the battery; wherein the second SOH curve characterizes the relationship between the battery's storage time and SOH; determining the aging parameters in the second SOH curve, and determining the second simulation data corresponding to the aging parameters, wherein the aging parameters include: the remaining time until the end of energy period (EOL).

[0023] In this embodiment, by using simulation tools and the electrochemical model to simulate the aging process of the battery, the SOH curve is obtained, and the second simulation data corresponding to the remaining time to the EOL point is determined. This enables the acquisition of accurate second simulation data and improves the accuracy of the first prediction model in predicting the remaining time to the EOL point.

[0024] In some embodiments, the charge / discharge process and the operating parameters are determined based on the process and parameters of the aging test performed on the battery.

[0025] In this embodiment, the charging and discharging process and operating parameters are determined according to the process and parameters of the aging test. This ensures that the generation conditions of the simulation data are consistent with the aging test conditions, reduces the error between the simulation data and the measured data, improves the reliability of the two data sources, and improves the prediction accuracy of the fusion prediction model for battery aging.

[0026] In some embodiments, the electrochemical model is determined based on the results of an aging pathway analysis performed on the battery; and the parameters of the electrochemical model are adjusted based on the results of a verification experiment performed on the battery.

[0027] In this embodiment, the electrochemical model is determined based on the aging pathway analysis and the model parameters are adjusted through verification experiments. This enables the electrochemical model to accurately reflect the actual aging mechanism of the battery. Further optimization of the model parameters through experimental verification can improve the authenticity and reliability of the simulation data and enhance the prediction accuracy of the first prediction model.

[0028] In some embodiments, obtaining first simulation data corresponding to the aging of the battery using the electrochemical model of the battery based on the battery's charge-discharge process and the operating parameters corresponding to the charge-discharge process includes: using simulation tools and the electrochemical model to simulate the aging process of the battery based on the charge-discharge process and the operating parameters to obtain the first simulation data.

[0029] In this embodiment, the battery aging process is efficiently simulated using simulation tools, which can quickly generate a large amount of simulation data across multiple scenarios, solving the problems of long acquisition cycles and high costs of actual test data, while also improving the reliability of the data.

[0030] In some embodiments, the electrochemical model includes a P2D model; the first aging prediction result and the second aging prediction result include the remaining time until the EOL point.

[0031] In this embodiment, the pseudo-two-dimensional model balances computational efficiency and simulation accuracy of the internal electrochemical processes of the battery, and uses the remaining time to the EOL point as the prediction result, so that the prediction result can meet the needs of battery life assessment, replacement planning and other requirements.

[0032] In some embodiments, according to a second aspect of this disclosure, a battery aging prediction device is provided, comprising: a simulation data acquisition module, configured to obtain first simulation data corresponding to the aging of the battery by using a simulation tool and an electrochemical model of the battery, based on a charge-discharge process of the battery and operating parameters corresponding to the charge-discharge process; wherein the electrochemical model includes a pseudo-two-dimensional (P2D) model; a first model construction module, configured to construct a first prediction model based on the first simulation data; a first prediction processing module, configured to obtain a first aging prediction result of the battery by using the trained first prediction model based on the first simulation data; a test data acquisition module, configured to perform an aging test on the battery to obtain first test data; a second model construction module, configured to construct a second prediction model based on the first test data; a prediction model fusion module, configured to fuse the first prediction model and the second prediction model to form a fused prediction model; and a second prediction processing module, configured to obtain a second aging prediction result of the battery by using the trained fused prediction model based on the first simulation data and the first test data.

[0033] In this embodiment, based on the battery's charging and discharging process and operating parameters, the battery's electrochemical model is used to obtain first simulation data. A first prediction model is then constructed based on this first simulation data, and the trained first prediction model is used to obtain aging prediction results. By utilizing the battery's charging and discharging process, operating parameters, and the characteristics of the electrochemical model to generate simulation data, the battery's aging model can be accurately simulated, and precise aging simulation data can be obtained. This allows the prediction model to learn the battery's aging patterns, improving the accuracy of battery aging predictions and enhancing the rationality, reliability, and versatility of battery aging predictions. Furthermore, by combining simulation data and test data and using a fusion prediction model for aging prediction, the large computational load of simulation and the potential bias in test data can be avoided, reducing prediction costs and improving prediction efficiency.

[0034] In some embodiments, according to a third aspect of this disclosure, an electronic device is provided, comprising: a memory; and a processor coupled to the memory, the processor being configured to perform the method described above based on instructions stored in the memory.

[0035] In some embodiments, according to a fourth aspect of this disclosure, a computer-readable storage medium is provided that stores computer instructions which are executed by a processor using the method described above.

[0036] In some embodiments, according to a fifth aspect of this disclosure, a computer program product is provided, the computer program product storing computer instructions which are executed by a processor using the method described above.

[0037] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0038] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0039] Figure 1 This is a schematic flowchart of some embodiments of the battery aging prediction method disclosed herein;

[0040] Figure 2 This is a schematic flowchart illustrating the determination of an electrochemical model in some embodiments of the battery aging prediction method of this disclosure.

[0041] Figure 3 This is a schematic diagram of the process of predicting aging through a fusion prediction model in some embodiments of the battery aging prediction method disclosed herein.

[0042] Figure 4 This is a schematic diagram illustrating the process of training a fusion prediction model in some embodiments of the battery aging prediction method disclosed herein.

[0043] Figure 5 This is a schematic diagram illustrating the process of constructing sample data for training a fusion prediction model in some embodiments of the battery aging prediction method of this disclosure.

[0044] Figure 6This is a schematic diagram of the process for obtaining second test data and aging parameters in some embodiments of the battery aging prediction method disclosed herein.

[0045] Figure 7 This is a schematic diagram of the first SOH curve in some embodiments of the battery aging prediction method of this disclosure;

[0046] Figure 8 This is a schematic diagram illustrating the process of obtaining second simulation data and aging parameters in some embodiments of the battery aging prediction method disclosed herein.

[0047] Figure 9 This is a schematic diagram of the second SOH curve in some embodiments of the battery aging prediction method of this disclosure;

[0048] Figure 10 This is a schematic diagram of the process for determining the value of the fusion loss function in some embodiments of the battery aging prediction method of this disclosure;

[0049] Figure 11 This is a schematic diagram of some embodiments of the battery aging prediction device disclosed herein;

[0050] Figure 12 Schematic diagrams of modules for some other embodiments of the battery aging prediction device of this disclosure;

[0051] Figure 13 This is a schematic diagram of modules for some embodiments of the electronic device disclosed herein. Detailed Implementation

[0052] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0053] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0054] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0055] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least some of the embodiments of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0056] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0057] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0058] In the description of the embodiments of this application, the technical terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of this application and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this application.

[0059] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0060] Figure 1 This is a flowchart illustrating some embodiments of the battery aging prediction method disclosed herein, such as... Figure 1 As shown, the battery aging prediction method includes steps S101 to S103:

[0061] Step S101: Based on the battery's charge and discharge process and the corresponding operating parameters, the first simulation data corresponding to the battery's aging is obtained using the battery's electrochemical model.

[0062] Batteries can be various types of rechargeable batteries, such as lithium batteries. The charging and discharging process includes constant current and constant voltage charging and constant current discharging, as well as information such as the charging and discharging time and intervals. Corresponding operating parameters include the battery's voltage, rate of change, and temperature during charging or discharging.

[0063] Electrochemical models can take many forms, such as the P2D model, which is a pseudo-two-dimensional model. This classic physicochemical model simulates the dynamic behavior of lithium-ion batteries by describing mass transfer, reaction kinetics, and charge distribution processes between the electrodes and electrolyte within the battery, based on mass and charge transfer. The P2D model mainly consists of a set of highly coupled partial differential equations (PDEs), balancing computational efficiency with the accuracy of simulating the internal electrochemical processes of the battery.

[0064] Battery aging can be categorized into calendar aging and other similar phenomena. Calendar aging refers to the gradual decline in battery performance during prolonged periods of non-use. The first simulation data corresponding to battery aging can be simulation data related to calendar aging, etc. This first simulation data includes various types of data, such as: time-related data, capacity-related data, lithium inventory-related data, active material-related data, electrode ratio-related data, and at least one of the following: time-related data includes simulated time; capacity-related data includes throughput capacity and state of equilibrium (SOH); lithium inventory-related data includes total lithium quantity, lithium inventory loss rate, and total lithium loss; active material-related data includes negative electrode active material loss rate and positive electrode active material loss rate; electrode ratio-related data includes positive / negative electrode ratio and actual positive / negative electrode capacity ratio; other data includes lithium concentration in the electrolyte, reaction rate constant, diffusion coefficient, and battery temperature.

[0065] Step S102: Construct a first prediction model based on the first simulation data.

[0066] The first prediction model can be a Transformer, Random Forest, Support Vector Machine, Multilayer Perceptron, Deep Neural Network, Recurrent Neural Network, Convolutional Neural Network, XGBoost (eXtreme Gradient Boosting), etc. For example, the first prediction model can be a multivariate temporal Transformer model, which is a neural network model based on a self-attention mechanism. Various methods can be used to construct a first prediction model that can predict battery aging based on the first simulation data. The input of the first prediction model is the first simulation data, and the output is the aging prediction result.

[0067] Step S103: Based on the first simulation data, obtain the first aging prediction result of the battery using the trained first prediction model.

[0068] The first prediction model can be trained using various training methods. The first aging prediction result can include various data such as battery capacity, voltage, SOH (State of Health), time to reach EOL (End of Life), and inflection point. SOH measures the degree of battery degradation and remaining lifespan, usually expressed as a percentage; EOL refers to the state where the capacity drops below a preset threshold (e.g., 70-80% SOH), failing to meet normal usage requirements. For example, the first aging prediction result includes the remaining time until the EOL point, which is the remaining time (in days, etc.) from the current time until the battery's EOL point. Using the remaining time until the EOL point as the prediction result allows the prediction result to meet the needs of battery life assessment and replacement planning.

[0069] The battery aging prediction method disclosed herein obtains first simulation data based on the battery's charge / discharge process and operating parameters, using the battery's electrochemical model. A first prediction model is then constructed based on this first simulation data, and the trained first prediction model is used to obtain aging prediction results. By utilizing the battery's charge / discharge process, operating parameters, and the characteristics of the electrochemical model to generate simulation data, the method can accurately simulate the battery's aging model and obtain precise aging simulation data. This allows the prediction model to learn the battery's aging patterns, improving the accuracy of battery aging predictions and enhancing the rationality, reliability, and versatility of the predictions. Furthermore, it avoids the need for battery aging tests, reducing prediction costs and improving prediction efficiency.

[0070] Figure 2 This is a schematic flowchart illustrating the determination of the electrochemical model in some embodiments of the battery aging prediction method of this disclosure, such as... Figure 2 As shown,

[0071] Step S201: Determine the electrochemical model based on the results of the aging pathway analysis of the battery.

[0072] Battery aging pathway analysis refers to the process of identifying the critical pathways and root causes of battery aging by studying the factors and manifestations of performance degradation during long-term use, storage, or cyclic charging and discharging. Various methods can be used for battery aging pathway analysis, and electrochemical models are determined based on the results.

[0073] For example, the aging process of lithium batteries can be described using mathematical formulas derived from fundamental electrochemical theorems and first principles. Multiple different aging pathways may exist simultaneously within a battery, and describing the complete battery aging pathway requires the use of multiple models, including kinetic and thermodynamic models. Kinetic models themselves are further divided into various types. For aging pathway analysis of batteries, a P2D model (which can be a fusion model or a multiphysics model) is determined to be used.

[0074] In the P2D model, mass transfer occurs at two locations: within the electrolyte (liquid phase) and inside the material particles (solid phase). Therefore, the mass transfer processes in both phases need to be analyzed separately. For the material particle interior (solid phase), Fick's second law is used to analyze the solid-phase diffusion coefficient (Ds) and the lithium concentration gradient (Cs) within the particles. The diffusion of lithium ions in the electrolyte (liquid phase) can be balanced using the surface concentration of the solid phase and the lithium ion exchange flux (j). The exchange flux can be determined by establishing a relationship with the exchange current, requiring the use of the Butler-Volmer equation at the electrolyte-electrode interface.

[0075] Based on the aging path analysis of the battery, and considering both computational efficiency and simulation accuracy, the Single Particle Model (SPM) was selected from the P2D model. SPM is a simplified physical or chemical model primarily used to describe the behavior or motion of individual particles (such as electrons, ions, and molecules), ignoring interactions between particles or other complex factors. It focuses on the state changes of individual particles (such as position, velocity, and energy), establishing a mathematical model through fundamental physical laws (such as Newton's laws of motion and quantum mechanics equations) to explain or predict particle behavior under specific conditions.

[0076] SPM is the fundamental "averaged" electrochemical model for lithium-ion batteries, which only considers solid-phase diffusion transport and spatially uniform kinetics. The SPM formula, describing the solid-phase diffusion equation, is shown below as an example:

[0077] (1-1);

[0078] Among them, c i (r,t) represents the concentration of the solid-phase diffused substance at radial position r and time t (i represents the solid-phase related parameter), which is time-dependent and is in mol / m³; Di is the solid-phase diffusion coefficient, in m² / s; r is the radial coordinate from the particle center to the surface, in m; t is the time, in seconds.

[0079] Taking the Di parameter as an example, let's analyze the relationship between temperature and material: the diffusion coefficient Di generally increases with increasing temperature; this relationship can be described by the Arrhenius equation; at the same time, Di is also related to the structure and composition of the electrode material, and different materials (such as graphite, silicon, lithium iron phosphate, etc.) have different diffusion coefficients. Therefore, for current batteries, it is necessary to specifically measure the parameters related to the material and specific temperature in all equations (not just the solid-phase diffusion equation here).

[0080] At the center of the sphere, due to symmetry, the concentration gradient must be zero; while at the surface of the sphere, the concentration gradient equals the molar flux j, which is related to the current density i on electrode i (here i can refer to either the positive or negative electrode) through the Faraday constant F and the number of electrons n participating in the reaction. Therefore, the boundary conditions are:

[0081] (1-2);

[0082] (1-3);

[0083] Where Ri is the particle radius (surface position); j i Let i be the molar flux (mass flow rate per unit area) of component i; i denoted as , where is the current density (current per unit area) for the reaction of component i; n is the number of electrons transferred in the electrode reaction; F is the Faraday constant; I is the total electrode current; Vi is the particle volume or the volume of the particle phase in the electrode; and ai is the specific surface area of ​​the particles.

[0084] The current density is related to the total battery current I by multiplying the electrode volume Vi by the effective electrode surface area ai. The specific surface area is a function of the particle radius Ri and the volume fraction of active material εi, with the specific relationship as follows:

[0085] (1-4);

[0086] Therefore, current density i i The overpotential ηi at battery temperature T can be calculated, and the specific relationship is as follows:

[0087] (1-5);

[0088] Where (1-6);

[0089] Among them, i i,0 is the exchange current density; ki is the reaction rate constant; c i (Ri,t) represents the component concentration on the particle surface; α represents the charge transfer coefficient; cel represents the component concentration in the electrolyte; and t represents time.

[0090] Since temperature is a key factor affecting battery degradation rate, a lumped thermal model was added to the SPM (Special Purpose Model). This thermal model includes three heat sources: one generated by the battery's DC resistance R... batt Ohmic heating caused by overpotential, reaction heating caused by entropy coefficient The entropy heat generated is determined by the cell's surface A. batt With a constant heat transfer coefficient h, heat is transferred to a temperature of T. env Cooling is achieved through convective heat transfer within the environment. The battery possesses a specific heat capacity cp, density ρ, and single-cell volume v. Therefore, the thermal model is as follows:

[0091] (1-7);

[0092] Where ρ is the battery density; V is the battery volume; c p I is the specific heat capacity at constant pressure; I is the operating current; R is the specific heat capacity at constant pressure. batt A is the battery internal resistance; h is the convective heat transfer coefficient; A batt T represents the battery surface area. env The ambient temperature.

[0093] Based on the Arrhenius equation and utilizing the diffusion activation energy E D,i and reaction rate activation energy E k,i From the reference temperature T ref The corresponding diffusion rate D iref Starting with the reference values ​​for the reaction rate Kiref, the relationship between these parameters and temperature was calculated:

[0094] (1-8);

[0095] (1-9);

[0096] Based on the results of the aging pathway analysis of the battery, an electrochemical model can be determined. Formulas 1-1 to 1-9 based on SPM can describe the physicochemical processes such as diffusion, reaction, and heat transfer inside the battery, providing a theoretical basis for generating simulation data through SPM. The dynamic behavior of the battery can be characterized from three dimensions: mass transport (diffusion), electrochemical reaction (current-overpotential), and heat transfer (temperature change). Combined with the temperature dependence equation, the influence of environmental factors on aging is reflected.

[0097] Step S202: Adjust the parameters of the electrochemical model based on the results of the verification experiment conducted on the battery.

[0098] Based on the analysis of battery aging pathways and the equations of the P2D model, the relevant parameters of the current battery can be preliminarily determined. Various methods can be used to determine which parameters of the P2D model need adjustment, allowing for targeted adjustments and improved fitting results. Validation experiments for the battery can be conducted using physical characterization tools, such as transmission electron microscopy, Raman spectroscopy, and X-ray photoelectron spectroscopy, to analyze the degradation mechanisms of the positive and negative electrodes. Based on the experimental validation results, parameters such as the solid-phase diffusion coefficient, liquid-phase diffusion coefficient (as needed), conductivity (as needed), and reaction activation energy can be adjusted to refine the electrochemical model.

[0099] By determining the electrochemical model based on aging pathway analysis and adjusting the model parameters through verification experiments, the electrochemical model can accurately reflect the actual aging mechanism of the battery. Further optimization of the model parameters through experimental verification can improve the authenticity and reliability of the simulation data and enhance the prediction accuracy of the first prediction model.

[0100] In some embodiments, based on the battery's charge-discharge process and corresponding operating parameters, a simulation tool and an electrochemical model are used to simulate the battery's aging process to obtain first simulation data. Various simulation tools can be used, such as PyBaMM, an open-source battery simulation framework based on Python that provides flexible battery models, solvers, and visualization capabilities, supporting battery R&D, engineering optimization, and educational applications. By combining the electrochemical model with PyBaMM, the battery's aging process is simulated based on the charge-discharge process and corresponding operating parameters to obtain the first simulation data.

[0101] By using simulation tools such as PyBaMM to efficiently simulate the battery aging process, a large amount of simulation data from multiple scenarios can be generated quickly, solving the problems of long acquisition cycles and high costs of actual test data, while improving the reliability of the data.

[0102] After the first prediction model is established, it can be trained using various methods. For example, historical simulation data can be obtained as samples for the first prediction model, and the actual prediction results corresponding to the samples can be used as the label values ​​of the samples. Training data can be constructed based on the samples and their corresponding label values, and the first prediction model can be trained independently using the training data.

[0103] The loss function for the first prediction model is set as follows:

[0104] (1-10);

[0105] Among them, MAE (Mean absolute error) is used to measure the average absolute error between the predicted value and the actual value of the first prediction model. The smaller the MAE, the better the model. It is the actual value, that is, the actual prediction result; is the predicted value output by the first prediction model, and n is the total number of samples.

[0106] Multiple model training methods can be used. The first prediction model is trained using the training data of the first prediction model to obtain the loss function value. The parameters of the first prediction model are adjusted according to the loss function value until the loss function value is less than the prediction threshold. Then the training stops, and the trained first prediction model is obtained.

[0107] Figure 3 The diagram below illustrates the process of predicting battery aging using a fusion prediction model in some embodiments of the battery aging prediction method disclosed herein. Figure 3 As shown,

[0108] Step S301: Perform an aging test on the battery to obtain the first test data.

[0109] The aging test performed on the battery can be a calendar aging test or other aging test. For example, a calendar aging test is performed on the battery to obtain first test data, which includes at least one of the following: SOH after storage, storage time, rate capability, temperature, coulombic efficiency / DC internal resistance, formulation, process, physicochemical properties, etc.

[0110] Step S302: Construct a second prediction model based on the first test data.

[0111] The second prediction model can be a Transformer, Random Forest, Support Vector Machine, Multilayer Perceptron, Deep Neural Network, Recurrent Neural Network, Convolutional Neural Network, etc. For example, the second prediction model can be a Transformer model. Various methods can be used to construct a second prediction model that can predict battery aging based on the first test data. The input to the second prediction model is the first test data, and the output aging prediction result of the second prediction model is the same as the output aging prediction result of the first prediction model. The second aging prediction result can include various data such as battery capacity, voltage, SOH, remaining time before EOL, and inflection point.

[0112] Step S303: The first prediction model and the second prediction model are fused to form a fused prediction model.

[0113] Step S304: Based on the first simulation data and the first test data, the second aging prediction result of the battery is obtained using the trained fusion prediction model.

[0114] Multiple methods can be used to fuse the first prediction model and the second prediction model to form a fused prediction model. The input of the fused prediction model is the first simulation data + the first test data. The output of the fused prediction model, the second aging prediction result, can be a weighted sum of the predicted values ​​of the first prediction model and the predicted values ​​of the second prediction model.

[0115] By combining simulation data and test data and using a fusion prediction model for aging prediction, we can avoid the large computational burden of simulating battery aging and the potential bias in test data obtained from battery aging tests. This avoids the limitations of a single model, improves the accuracy of battery aging prediction, reduces prediction costs, and increases prediction efficiency.

[0116] Before using the trained fusion prediction model to obtain the second aging prediction results of the battery, the fusion prediction model is trained. Figure 4 The diagram below illustrates the process of training the fusion prediction model in some embodiments of the battery aging prediction method disclosed herein, such as... Figure 4 As shown,

[0117] Step S401: Construct sample data for training the fusion prediction model. Various methods can be used to construct the sample data for training the fusion prediction model.

[0118] Step S402: Train the fusion prediction model using sample data, and adjust the parameters of the fusion prediction model according to the fusion loss function value of the fusion prediction model to obtain the trained fusion prediction model.

[0119] By constructing sample data and training the fusion prediction model based on the fusion loss function, the fusion prediction model can fully learn the effective information from simulation data and test data, enabling the trained fusion prediction model to have better generalization ability and improve the prediction accuracy of battery aging.

[0120] In some embodiments, the first model sample data includes a first sample and a first label value corresponding to the first sample, and the second model sample data includes a second sample and a second label value corresponding to the second sample. Figure 5 The diagram illustrates the process of constructing sample data for training the fusion prediction model in some embodiments of the battery aging prediction method of this disclosure, such as... Figure 5 As shown,

[0121] Step S501: Based on the charging and discharging process and operating parameters, the second simulation data and aging parameters corresponding to the aging of the battery are obtained using an electrochemical model.

[0122] Step S502: Based on the second simulation data and aging parameters, construct the first model sample data, wherein the first sample is the second simulation data and the first label value is the aging parameter.

[0123] Step S503: Perform an aging test on the battery to obtain second test data and aging parameters.

[0124] Step S504: Based on the second test data and aging parameters, construct second model sample data, where the second sample is the second test data and the second label value is the aging parameter.

[0125] The electrochemical model can be a P2D model, and the aging parameters can be the remaining time until the end of energy (EOL). The charge / discharge process, operating parameters, and electrochemical model used to obtain the second simulation data can be the same as those used to obtain the first simulation data. Similarly, the aging test process and corresponding parameters used to obtain the second test data can be the same as those used to obtain the first test data. Multiple methods can be used to construct the first model sample data and the second model sample data.

[0126] Step S505: Construct sample data based on the first model sample data and the second model sample data.

[0127] Sample data can be constructed using various methods. For example, the first sample and the second sample of the first model sample data corresponding to the aging parameter (label value) can be combined to generate the model sample of the sample data; the weighted sum of the aging parameter of the first model sample data and the aging parameter of the second model sample data corresponding to this aging parameter (label value) can be used as the label value of the sample data.

[0128] In this embodiment, sample data is constructed based on the second simulation data generated by the electrochemical model, the second test data obtained by the aging test, and the corresponding aging parameters. This sample data includes simulation data that reflects the theoretical laws of battery aging and test data that reflects the aging characteristics of the battery in actual use. This provides high-quality and comprehensive samples for training the fusion prediction model and improves the prediction accuracy of the fusion prediction model for battery aging.

[0129] Based on the second simulation data generated by the electrochemical model, the second test data obtained from the aging test, and the corresponding aging parameters, sample data is constructed. This sample data includes simulation data that reflects the theoretical laws of battery aging and test data that reflects the actual aging characteristics of the battery. It can provide high-quality and comprehensive samples for training the fusion prediction model and improve the prediction accuracy of the fusion prediction model for battery aging.

[0130] Figure 6 This is a schematic diagram illustrating the process of obtaining second test data and aging parameters in some embodiments of the battery aging prediction method of this disclosure, such as... Figure 6 As shown,

[0131] Step S601: Perform an aging test on the battery to obtain the first state of health (SOH) curve of the battery, wherein the first SOH curve characterizes the relationship between the storage time of the battery and the SOH.

[0132] Step S602: Determine the second test data corresponding to the aging parameters based on the first SOH curve; wherein, the aging parameters include: the remaining time until the EOL point.

[0133] In some embodiments, the aging parameters can be multiple parameters, such as the remaining time until the EOL point, i.e., the aging parameter is DTE (Day to EOL), and EOL can be set to 70% SOH. After battery storage, a calendar aging test is performed on the battery. The operator performs two cycles of charge-discharge testing on the battery every 7 days and records the data. The maximum discharge capacity of the second cycle test is represented by DQ, DQ = max(abs(Qs)), where Qs is the capacity sequence of the discharge stage in the second cycle test, abs() is the absolute value function, and max() is the maximum value function. The length of the capacity sequence depends on the battery's rated capacity, discharge current, discharge duration, sampling frequency, etc. Based on the cycle sequence stored each time, a discharge capacity DQ can be determined, and the ratio of DQ to D0 is taken as SOH, where D0 represents the initial discharge capacity.

[0134] Based on all collected discrete SOH data throughout the storage lifecycle, a first SOH curve is generated. This first SOH curve represents the SOH curve of the calendar-aged battery. Figure 7As shown, the first SOH curve characterizes the relationship between battery storage time and SOH. Figure 7 The horizontal axis represents the number of days stored, and the vertical axis represents the SOH value.

[0135] In addition to the stored state of equilibrium (SOH), the DC internal resistance, coulombic efficiency, and capacity differential value corresponding to the second charge-discharge cycle were calculated during the test. Multiple stored measurements can be stitched together into a sequence (curve). These characteristics are affected by different battery formulations, data quality, and storage methods. DC internal resistance: The DC internal resistance value is obtained by comparing the voltage difference between the fully charged and resting state and the first 30 seconds of discharge; it reflects the battery's performance state and charge-discharge characteristics. Coulombic efficiency is the ratio of charge capacity to discharge capacity, reflecting the proportional relationship between the energy stored and the energy actually released during discharge. The capacity increment curve is the relationship between dQ (capacity increment) and dV (voltage increment) versus voltage, reflecting the rate of change in the system's ability to release charge under a given voltage change. The X-axis represents voltage, and the Y-axis represents the ratio.

[0136] Based on the first SOH curve, the second test data corresponding to the aging parameters can be determined. For example, the DTE can be determined from the first SOH curve, and the second test data corresponding to the DTE can be determined. Based on the correspondence between SOH and DTE in the first SOH curve obtained through the first SOH curve, and the corresponding data obtained in the calendar aging test, such as SOH, storage time, rate capability, temperature, coulombic efficiency / DC internal resistance, formulation, process, and physicochemical properties, a test dataset is generated. The test dataset can be a two-dimensional array. The test dataset is shown in Table 1 below:

[0137]

[0138] Table 1 - Battery Test Data Set

[0139] Several methods can be used to obtain second test data corresponding to aging parameters from the test dataset. Multiple methods can be used, with DTE as the prediction label, to generate second test data by sequentially sliding the data through a pre-set sliding window of length in the two-dimensional array in Table 1. For example, the two-dimensional array corresponding to the test datasets of all batteries can be traversed, using DTE as the label for the regression algorithm, and a time-series dataset with a sliding window length of 3 can be generated as the second test data. The second test data includes SOH, storage time, rate capability, temperature, coulombic efficiency / DC internal resistance, formulation, process, and physicochemical properties after storage testing. The generated second model sample data includes a second sample and a corresponding second label value. The second sample can include SOH, storage time, rate capability, temperature, coulombic efficiency / DC internal resistance, formulation, process, and physicochemical properties after storage testing, and the second label value is DTE.

[0140] By using the SOH curve obtained from battery aging tests, a second test data corresponding to the remaining time to the EOL point can be determined. This provides accurate second test data and improves the accuracy of the second prediction model in predicting the remaining time to the EOL point.

[0141] In some embodiments, the charging and discharging process and operating parameters are determined based on the process and parameters of the battery aging test. The battery aging test can be a calendar aging test or a cycle aging test, etc. The process of the battery aging test includes charging and discharging the battery, as well as the corresponding charging and discharging process time, charging and discharging interval time, charging and discharging cycle number, and other information; the aging test parameters include: battery voltage, rate, temperature, and other parameters during charging or discharging.

[0142] Determining the charging and discharging process and operating parameters based on the aging test process and parameters can ensure that the generation conditions of simulation data are consistent with the aging test conditions, reduce the error between simulation data and measured data, improve the reliability of the two data sources, and improve the prediction accuracy of the fusion prediction model for battery aging.

[0143] Figure 8 The diagram below illustrates the process of obtaining second simulation data and aging parameters in some embodiments of the battery aging prediction method disclosed herein, such as... Figure 8 As shown:

[0144] Step S801: Based on the battery's charging and discharging process and operating parameters, use simulation tools and an electrochemical model to simulate the battery's aging process and obtain the battery's second state of health (SOH) curve; wherein, the second SOH curve characterizes the relationship between the battery's storage time and SOH.

[0145] Step S802: Based on the second SOH curve, determine the second simulation data corresponding to the aging parameters, wherein the aging parameters include: the remaining time until the EOL point.

[0146] In some embodiments, the aging parameter includes the remaining time until the EOL point, i.e., the aging parameter is DTE (Day to EOL), and EOL can be set to 70% SOH. Based on the process and parameters for calendar aging testing of the battery, the battery's charge / discharge process and corresponding operating parameters are determined. Based on the determined charge / discharge process and operating parameters, the SPM model is simulated using the simulation tool PyBaMM to obtain simulation data; wherein, the simulation data includes the battery's second state of health (SOH) curve, such as... Figure 9 As shown, the second SOH curve characterizes the relationship between the battery's storage time and SOH.

[0147] The second simulation data corresponding to the aging parameters can be determined based on the second SOH curve. For example, the DTE can be determined from the second SOH curve, and the corresponding second simulation data can be determined. The correspondence between SOH and DTE in the second SOH curve is obtained through the second SOH curve, and a test dataset is generated using the data output by the simulation tool PyBaMM. The test dataset can be a two-dimensional array. The data output by PyBaMM includes: time-related data, capacity-related data, lithium inventory-related data, active material-related data, electrode ratio-related data, and other data. Each test data in the test dataset includes: time-related data, capacity-related data, lithium inventory-related data, active material-related data, electrode ratio-related data, other data, and DTE. Various methods can be used to determine the second test data corresponding to the aging parameters. The second test data can be generated by using DTE as the prediction label and sliding the data sequentially within the two-dimensional array corresponding to the test dataset according to a pre-set sliding window length.

[0148] For example, a two-dimensional array corresponding to the test datasets of all batteries can be traversed, using DTE as the label for the regression algorithm. A time-series dataset of length 3 is generated as the second simulation data. The second simulation data includes time-related data, capacity-related data, lithium inventory-related data, active material-related data, electrode ratio-related data, and other data. The generated first model sample data includes a first sample and a first label value corresponding to the first sample. The first sample may include time-related data, capacity-related data, lithium inventory-related data, active material-related data, electrode ratio-related data, and other data. The first label value is DTE.

[0149] By using simulation tools and electrochemical models to simulate the aging process of batteries, the SOH curve is obtained, and second simulation data corresponding to the remaining time to the EOL point is determined. This allows for the acquisition of accurate second simulation data, which can improve the accuracy of the first prediction model in predicting the remaining time to the EOL point.

[0150] Figure 10 This is a flowchart illustrating the process of determining the fusion loss function value in some embodiments of the battery aging prediction method of this disclosure, such as... Figure 10 As shown,

[0151] Step S1001: Based on the sample prediction values ​​of the first prediction model and the sample prediction values ​​of the second prediction model, determine the predicted average value and obtain the corresponding sample label value.

[0152] For example, the sample data used to train the fusion prediction model includes: data A, etc.; data A includes model sample A and sample label value A; model sample A is composed of the first sample B of the first model sample data and the second sample C of the second model sample data, and the sample label value A is the weighted sum of the first label value D of the corresponding first model sample data and the second label value E of the second model sample data, and the first label value D and the second label value E can be the same or different.

[0153] When training the fusion prediction model, model sample A is input into the fusion prediction model to obtain the sample prediction value of the first prediction model output based on the first sample B, and the sample prediction value of the second prediction model output based on the second sample C. The corresponding sample label value A is obtained based on the sample prediction values ​​of the first and second prediction models. Multiple methods can be used to determine the average prediction value based on the sample prediction values ​​of the first and second prediction models.

[0154] Step S1002: Determine the absolute value of the prediction difference based on the predicted average value and the sample label value.

[0155] Step S1003: Use the sum of the absolute value of the prediction difference and the prediction penalty value as the fusion loss function value, or use the absolute value of the prediction difference as the fusion loss function value.

[0156] The fusion loss function value is determined by combining the absolute value of the difference between the predicted average and the sample label value with the prediction penalty value. Using the absolute value of the difference can determine the deviation between the predicted value and the true value, improving the sensitivity of the fusion loss function to the prediction error. Using the prediction penalty value can avoid extreme prediction deviations, enabling the fusion prediction model to converge towards a better direction, and improving the prediction accuracy and stability of battery aging.

[0157] In some embodiments, the predicted penalty value includes a first penalty value; the absolute value of the first difference is determined based on the sample predicted value of the first prediction model and the sample predicted value of the second prediction model; and the absolute value of the first difference is weighted by a first weighting parameter to obtain the first penalty value.

[0158] For example, the fusion loss function is:

[0159] (1-11);

[0160] in, These are the sample predictions from the first prediction model. These are the sample predictions from the second prediction model. y represents the predicted average; y is the corresponding sample label value; To predict the absolute value of the difference; Let λ be the absolute value of the first difference, and λ be the first weighting parameter. This is the first penalty value.

[0161] By minimizing the prediction difference between the two models, the residual is the difference between the average of the two predictions and the label. The absolute value of the first difference is determined based on the sample predictions of the first and second prediction models. The first weighting parameter is used for weighting to obtain the first penalty value. The penalty value can be determined based on the prediction difference between the first and second prediction models and the weighting parameter, which can make the predictions of the first and second prediction models more consistent, reduce prediction conflicts between models, strengthen the coordination between models, and improve the prediction accuracy and prediction stability of battery aging.

[0162] In some embodiments, the predicted penalty value includes a second penalty value; a second absolute value of the difference is determined based on the sample predicted value and the sample label value of the first prediction model; a third absolute value of the difference is determined based on the sample predicted value and the sample label value of the second prediction model; and a second penalty value is obtained by weighting the sum of the second absolute value of the difference and the third absolute value of the difference using a second weighting parameter.

[0163] For example, the fusion loss function is:

[0164] (1-12);

[0165] in, To predict the absolute value of the difference, The absolute value of the second difference. The absolute value of the third difference; The second weighting parameter, This is the second penalty value.

[0166] Considering the potential competition between the two models, a second weighting parameter is used to weight the samples based on the absolute difference between the predicted values ​​of the first and second prediction models and the sample label values, respectively, to obtain a second penalty value. This penalty value can be determined based on the absolute difference between the predicted values ​​of the first and second prediction models and the sample label values, as well as the weighting parameter. This can simultaneously constrain the deviation of the first and second prediction models from the true values, avoid a single model dominating the prediction results, and ensure that the first and second prediction models have independent prediction accuracy. This can improve the prediction accuracy and stability of battery aging.

[0167] In some embodiments, the fusion loss function is:

[0168] (1-13);

[0169] Based on Equation 1-13, the absolute value of the prediction difference can be used as the value of the fusion loss function without considering the prediction penalty value, which can improve the efficiency of training.

[0170] In some embodiments, the sample prediction values ​​of the first prediction model are weighted by a third weighting parameter to obtain a first prediction weighted value; the sample prediction values ​​of the second prediction model are weighted by a fourth weighting parameter to obtain a second prediction weighted value; the sum of the third and fourth weighting parameters is 1; a prediction weighted sum is determined based on the first and second prediction weighted values; and the absolute value of the prediction weighted difference is determined based on the prediction weighted sum and sample label values, which serves as the fusion loss function value. For example, the fusion loss function is:

[0171] (1-14);

[0172] in, As the third weighting parameter, It is the fourth weighting parameter; The first prediction weighting value, This is the second prediction weighting value; To predict the weighted sum, To predict the absolute value of the weighted difference.

[0173] By weighting the predicted values ​​of the first and second prediction models respectively using weighting parameters, and calculating the absolute value of the difference between the sum of the weighted results and the sample label values, the fusion loss function can be used to dynamically adjust the contribution of the first and second prediction models to the fusion prediction. This allows the fusion prediction model to adaptively optimize the weights, improves the guidance of the fusion loss function for training the fusion prediction model, and enhances the prediction accuracy of battery aging.

[0174] The fusion prediction model is trained based on the above fusion loss function. During training, the core objective of residual calibration is to quantify and correct the deviation between the model prediction and the true label. The calibration objective is to identify the residual distribution of the two models on the dataset (such as mean, variance, and deviation direction) and correct the predicted value to make it closer to the true label.

[0175] After the first prediction model is trained, based on the same battery charge-discharge process and corresponding operating parameters as the second simulation data, as well as the electrochemical model, the first simulation data corresponding to battery aging is obtained. The prediction data is then input into the trained first prediction model to obtain the first aging prediction result, which includes the remaining time until the end of energy period (EOL).

[0176] After the fusion model is trained, based on the same battery charge-discharge process and corresponding operating parameters as the second simulation data, as well as the electrochemical model, the first simulation data corresponding to battery aging is obtained. The battery is then subjected to the same aging test as the second test data to obtain the first test data. Based on the first and second simulation data, prediction data is generated and input into the trained fusion prediction model to obtain the second aging prediction result, which includes the remaining time until the end of energy period (EOL).

[0177] In some embodiments, such as Figure 11 As shown, this disclosure provides a battery aging prediction device, including a simulation data acquisition module 1101, a first model construction module 1102, and a first prediction processing module 1103. The simulation data acquisition module 1101 obtains first simulation data corresponding to battery aging based on the battery's charge-discharge process and corresponding operating parameters, using the battery's electrochemical model. The first model construction module 1102 constructs a first prediction model based on the first simulation data. The first prediction processing module 1103 obtains the first aging prediction result of the battery using the trained first prediction model based on the first simulation data.

[0178] In some embodiments, such as Figure 12 As shown, in addition to the simulation data acquisition module 1101, the first model construction module 1102, and the first prediction processing module 1103, the battery aging prediction device also includes: a test data acquisition module 1104, a second model construction module 1105, a prediction model fusion module 1106, a second prediction processing module 1107, a sample construction module 1108, a fusion model training module 1109, and a loss determination module 1110.

[0179] The test data acquisition module 1104 performs an aging test on the battery to obtain first test data. The second model construction module 1105 constructs a second prediction model based on the first test data. The prediction model fusion module 1106 fuses the first and second prediction models to form a fused prediction model. The second prediction processing module 1107 obtains the second aging prediction result of the battery using the trained fused prediction model based on the first simulation data and the first test data.

[0180] The sample construction module 1108 constructs sample data for training the fusion prediction model. The fusion model training module 1109 uses the sample data to train the fusion prediction model and adjusts the parameters of the fusion prediction model according to the fusion loss function value of the fusion prediction model to obtain the trained fusion prediction model.

[0181] The loss determination module 1110 determines the predicted average value and obtains the corresponding sample label value based on the sample predicted values ​​of the first prediction model and the second prediction model; the loss determination module 1110 determines the absolute value of the prediction difference based on the predicted average value and the sample label value; the loss determination module 1110 uses the sum of the absolute value of the prediction difference and the prediction penalty value as the fusion loss function value, or uses the absolute value of the prediction difference as the fusion loss function value.

[0182] In some embodiments, the loss determination module 1110 determines the absolute value of the first difference based on the sample predicted values ​​of the first prediction model and the sample predicted values ​​of the second prediction model; the loss determination module 1110 performs weighted processing on the absolute value of the first difference using a first weighting parameter to obtain a first penalty value.

[0183] The loss determination module 1110 determines the second absolute value of the difference based on the sample predicted value and sample label value of the first prediction model; the loss determination module 1110 determines the third absolute value of the difference based on the sample predicted value and sample label value of the second prediction model; the loss determination module 1110 performs weighted processing on the sum of the second absolute value of the difference and the third absolute value of the difference using the second weighting parameter to obtain the second penalty value.

[0184] The loss determination module 1110 weights the sample prediction values ​​of the first prediction model using a third weighting parameter to obtain a first prediction weighted value; the loss determination module 1110 weights the sample prediction values ​​of the second prediction model using a fourth weighting parameter to obtain a second prediction weighted value; wherein the sum of the third weighting parameter and the fourth weighting parameter is 1; the loss determination module 1110 determines the prediction weighted sum based on the first prediction weighted value and the second prediction weighted value; the loss determination module 1110 determines the absolute value of the prediction weighted difference based on the prediction weighted sum and sample label values, which is used as the fusion loss function value.

[0185] In some embodiments, the sample construction module 1108 obtains second simulation data and aging parameters corresponding to battery aging using an electrochemical model based on the charge / discharge process and operating parameters; the sample construction module 1108 constructs first model sample data based on the second simulation data and aging parameters, wherein the first sample is the second simulation data and the first label value is the aging parameter; the sample construction module 1108 performs aging tests on the battery to obtain second test data and aging parameters; the sample construction module 1108 constructs second model sample data based on the second test data and aging parameters, wherein the second sample is the second test data and the second label value is the aging parameter; the sample construction module 1108 constructs sample data based on the first model sample data and the second model sample data.

[0186] The sample construction module 1108 performs an aging test on the battery to obtain the first state of health (SOH) curve of the battery. The first SOH curve represents the relationship between the storage time of the battery and the SOH. The sample construction module 1108 determines the aging parameters from the first SOH curve and determines the second test data corresponding to the aging parameters. The aging parameters include the remaining time until the end of the energy period (EOL).

[0187] The sample construction module 1108 uses simulation tools and an electrochemical model to simulate the aging process of the battery based on the charging and discharging process and operating parameters, and obtains the second state of health (SOH) curve of the battery. The second SOH curve represents the relationship between the battery's storage time and SOH. The sample construction module 1108 determines the aging parameters from the second SOH curve and determines the second simulation data corresponding to the aging parameters. The aging parameters include the remaining time until the end of the energy period (EOL).

[0188] The simulation data acquisition module 1101 determines the charge / discharge process and operating parameters based on the aging test procedures and parameters performed on the battery. Based on the results of the aging path analysis of the battery, the simulation data acquisition module 1101 determines the electrochemical model; based on the results of the verification experiments performed on the battery, the simulation data acquisition module 1101 adjusts the parameters of the electrochemical model. Based on the charge / discharge process and operating parameters, the simulation data acquisition module 1101 uses simulation tools and the electrochemical model to simulate the aging process of the battery, obtaining the first set of simulation data.

[0189] Figure 13 This is a schematic diagram of modules according to some embodiments of an electronic device based on the present disclosure. For example... Figure 13 As shown, the electronic device may include a memory 1301, a processor 1302, a communication interface 1303, and a bus 1304. The memory 1301 is used to store instructions, and the processor 1302 is coupled to the memory 1301. The processor 1302 is configured to execute the battery aging prediction method described above based on the instructions stored in the memory 1301.

[0190] Memory 1301 can be high-speed RAM, non-volatile memory, etc., and memory 801 can also be a memory array. Memory 1501 may also be divided into blocks, and the blocks can be combined into virtual volumes according to certain rules. Processor 1302 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the battery aging prediction method of this disclosure.

[0191] In some embodiments, this disclosure provides a computer-readable storage medium storing computer instructions that are executed by a processor as in any of the above embodiments for predicting battery aging.

[0192] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples (not an exhaustive list) of readable storage media may include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0193] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0194] Embodiments of this disclosure may also be computer program products, including computer program instructions that, when executed by a processor, cause the processor to perform the steps in the battery aging prediction methods according to various embodiments of this disclosure as described in the "Exemplary Methods" section above.

[0195] The steps of the methods disclosed herein are not limited to the specific order described above, unless otherwise specifically stated. Furthermore, in some embodiments, this disclosure may also be implemented as a program recorded on a recording medium, the program including machine-readable instructions for implementing the methods according to this disclosure. Therefore, this disclosure also covers recording media storing programs for performing the methods according to this disclosure.

[0196] Although this application has been described with reference to preferred embodiments, various modifications can be made thereto and components can be replaced with equivalents without departing from the scope of this application. In particular, the technical features mentioned in the various embodiments can be combined in any manner, provided there is no structural conflict. This application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for predicting battery aging, characterized in that, include: Based on the battery's charge-discharge process and the corresponding operating parameters, simulation tools are used and the battery's electrochemical model is utilized to obtain first simulation data corresponding to the battery's aging; wherein, the electrochemical model includes a pseudo-two-dimensional model (P2D model). Based on the first simulation data, a first prediction model is constructed; Based on the first simulation data, the first aging prediction result of the battery is obtained using the trained first prediction model; An aging test is performed on the battery to obtain first test data; Based on the first test data, construct a second prediction model; The first prediction model and the second prediction model are fused to form a fused prediction model; Based on the first simulation data and the first test data, the second aging prediction result of the battery is obtained using the trained fusion prediction model; The method includes, before obtaining the second aging prediction result of the battery using the trained fusion prediction model, the following steps: Construct sample data for training the fusion prediction model; The fusion prediction model is trained using the sample data, and the parameters of the fusion prediction model are adjusted according to the fusion loss function value of the fusion prediction model to obtain the trained fusion prediction model. Specifically, based on the sample predicted values ​​of the first prediction model and the sample predicted values ​​of the second prediction model, a predicted average value is determined and a corresponding sample label value is obtained; based on the predicted average value and the sample label value, an absolute value of the prediction difference is determined; the sum of the absolute value of the prediction difference and the prediction penalty value is used as the fusion loss function value, or the absolute value of the prediction difference is used as the fusion loss function value.

2. The battery aging prediction method as described in claim 1, characterized in that, The predicted penalty value includes: a first penalty value; the method includes: The absolute value of the first difference is determined based on the sample predicted values ​​of the first prediction model and the sample predicted values ​​of the second prediction model. The first penalty value is obtained by weighting the absolute value of the first difference using the first weighting parameter.

3. The battery aging prediction method as described in claim 1, characterized in that, The predicted penalty value includes: a second penalty value; the method includes: The absolute value of the second difference is determined based on the sample prediction value of the first prediction model and the sample label value. The absolute value of the third difference is determined based on the sample predicted value of the second prediction model and the sample label value. The second penalty value is obtained by weighting the sum of the absolute values ​​of the second and third differences using a second weighting parameter.

4. The battery aging prediction method as described in claim 1, characterized in that, The method includes: The sample prediction values ​​of the first prediction model are weighted by a third weighting parameter to obtain a first prediction weighting value. The sample prediction values ​​of the second prediction model are weighted by a fourth weighting parameter to obtain a second prediction weighted value; wherein the sum of the third weighting parameter and the fourth weighting parameter is 1. Based on the first predicted weighted value and the second predicted weighted value, determine the predicted weighted sum and obtain the corresponding sample label value; The absolute value of the prediction weighted sum and the sample label value are determined as the value of the fusion loss function.

5. The battery aging prediction method as described in claim 1, characterized in that, The first model sample data includes a first sample and a first label value corresponding to the first sample; the second model sample data includes a second sample and a second label value corresponding to the second sample. The sample data used to train the fusion prediction model includes: Based on the charging and discharging process and the operating parameters, the electrochemical model is used to obtain second simulation data and aging parameters corresponding to the aging of the battery. Based on the second simulation data and the aging parameters, the first model sample data is constructed, wherein the first sample is the second simulation data and the first label value is the aging parameters; The battery is subjected to the aging test to obtain second test data and the aging parameters. Based on the second test data and the aging parameters, the second model sample data is constructed, wherein the second sample is the second test data and the second label value is the aging parameters; The sample data is constructed based on the first model sample data and the second model sample data.

6. The battery aging prediction method as described in claim 5, characterized in that, The step of performing the aging test on the battery to obtain the second test data and the aging parameters includes: An aging test is performed on the battery to obtain a first SOH curve of the battery, wherein the first SOH curve characterizes the relationship between the storage time of the battery and SOH. The aging parameters are determined in the first SOH curve, and the second test data corresponding to the aging parameters are determined, wherein the aging parameters include: the remaining time until the end of battery life (EOL).

7. The battery aging prediction method as described in claim 5, characterized in that, The step of obtaining second simulation data and aging parameters corresponding to the aging of the battery using the electrochemical model based on the charging and discharging process and the operating parameters includes: Based on the charging and discharging process and the operating parameters, the aging process of the battery is simulated using simulation tools and the electrochemical model to obtain the second SOH curve of the battery; wherein, the second SOH curve characterizes the relationship between the storage time of the battery and SOH. The aging parameters are determined in the second SOH curve, and the second simulation data corresponding to the aging parameters are determined, wherein the aging parameters include: the remaining time until the EOL point.

8. The battery aging prediction method as described in claim 1, characterized in that, include: The charging and discharging process and the operating parameters are determined based on the aging test procedure and parameters performed on the battery.

9. The battery aging prediction method as described in claim 1, characterized in that, include: The electrochemical model was determined based on the results of the aging pathway analysis performed on the battery. The parameters of the electrochemical model were adjusted based on the results of the verification experiments conducted on the battery.

10. The battery aging prediction method according to any one of claims 1 to 9, characterized in that, The first aging prediction result and the second aging prediction result include: the remaining time until the EOL point.

11. A battery aging prediction device, characterized in that, include: The simulation data acquisition module is used to obtain first simulation data corresponding to the aging of the battery by using simulation tools and the electrochemical model of the battery, based on the battery's charge and discharge process and the corresponding operating parameters; wherein the electrochemical model includes a pseudo-two-dimensional model (P2D model). The first model building module is used to build a first prediction model based on the first simulation data; The first prediction processing module is used to obtain the first aging prediction result of the battery based on the first simulation data and the trained first prediction model. The test data acquisition module is used to perform an aging test on the battery and obtain first test data; The second model building module is used to build a second prediction model based on the first test data; The prediction model fusion module is used to fuse the first prediction model and the second prediction model to form a fused prediction model; The second prediction processing module is used to obtain the second aging prediction result of the battery based on the first simulation data and the first test data using the trained fusion prediction model. A sample construction module is used to construct sample data for training the fusion prediction model; The fusion model training module is used to train the fusion prediction model using the sample data and adjust the parameters of the fusion prediction model according to the fusion loss function value of the fusion prediction model to obtain the trained fusion prediction model. The loss determination module is used to determine the prediction average and obtain the corresponding sample label value based on the sample prediction values ​​of the first prediction model and the sample prediction values ​​of the second prediction model; determine the absolute value of the prediction difference based on the prediction average and the sample label value; and use the sum of the absolute value of the prediction difference and the prediction penalty value as the fusion loss function value, or use the absolute value of the prediction difference as the fusion loss function value.

12. An electronic device, characterized in that, include: Memory; And a processor coupled to the memory, the processor being configured to perform the method as described in any one of claims 1 to 10 based on instructions stored in the memory.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are executed by a processor according to any one of claims 1 to 10.

14. A computer program product, characterized in that, The computer program product stores computer instructions that are executed by a processor using the method as described in any one of claims 1 to 10.

Images