Power battery design and management method based on digital twinning

Abstract

A power battery design and management method based on digital twinning includes steps of establishing a multi-physical field coupling mechanism model mapped with a battery sample, calibrating the multi-physical field coupling mechanism model by using experimental data, performing sensitivity analysis on a calculation result data set of the multi-physical field mechanism model by using an active subspace method, screening out key parameters affecting battery performance as main control parameters of the model, and establishing a neural network high-efficiency quantitative simulation model, which can realize a power battery twinning model mapping a physical entity, etc.The method can reduce cost, realize agile development, reduce workload and blindness of parameter identification, has high precision and stability, and can also realize full-life cycle closed-loop control of the power battery.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automation, to the field of power battery design and modeling, and in particular to a power battery design and management method based on digital twinning. BACKGROUND

[0002] China's new energy vehicles have entered a critical period of large-scale application. As a key node in the "three vertical and three horizontal" research and development layout, designing high-strength, lightweight, high-safety, low-cost, and long-life power batteries and management systems has become a key problem. With the use of electric vehicles, harsh road conditions, environmental temperature, and dynamic changes in load can cause the performance of the battery system to decline nonlinearly, and in turn cause problems such as liquid leakage, insulation damage, and partial short circuits. If the internal reaction mechanism of the battery is not clear, the failure characteristics are not monitored in a timely manner, and the health status is not assessed, the battery will accelerate aging and cause serious safety accidents such as self-ignition and explosion. Therefore, it is of great significance to achieve agile design and efficient management of power batteries.

[0003] Digital twinning technology is a simulation process that fully utilizes physical models, sensor updates, operation history, and other data, integrates multi-disciplinary, multi-physical, multi-scale, and multi-probability, and completes mapping in a virtual space, thereby reflecting the full life cycle process of the corresponding entity equipment. Digital twinning is a concept beyond reality, and can be regarded as a digital mapping system of one or more important, mutually dependent, and coupled systems.

[0004] At present, for the agile design and management of power battery, the invention patent with publication number CN 111581850 A discloses a full-cycle power battery management system applying digital twin technology, which specifically discloses that the system is composed of a physical model system, a physical management system and a cloud twin system; wherein, the physical model system is based on the physical model of real battery monomer, module and power battery system, as the controlled object of the physical management system, and as the data source of the battery management algorithm of the physical management system, and as the data source of the twin model of the cloud twin system; the physical management system obtains the initial parameters generated in the development process of the physical model system and the operation data generated in the use process through the data acquisition module, processes these data, calculates the running battery management algorithm, and uploads the data from the physical model system to the cloud twin system; the cloud twin system constructs a battery twin model based on the data information uploaded by the physical management system and based on the cloud algorithm, and performs data tracing, state estimation, safety diagnosis, life prediction and working condition analysis functions through the battery twin model and battery historical data; part of the output information flow of the cloud twin system flows to the physical management system, providing reference basis for state monitoring, decision making, control execution and the like of the physical management system, and finally realizing control of the physical model system through the physical management system; another part directly flows to the physical model system, providing reference and guidance for battery material modification and structure optimization, and providing reference and guidance for the battery management algorithm development process, for classification and reorganization of batteries in different aging states, and realizing battery reorganization and cascade utilization; in the battery management algorithm and the cloud algorithm, the battery management algorithm adopts a relatively short time scale algorithm, and the cloud algorithm adopts a relatively long time scale algorithm, which can effectively reduce the experimental amount in the early development process and shorten the development cycle compared with the prior art. The invention patent with publication number CN110534823B discloses a power battery equalization management system and method, which establishes a power battery equalization management system based on cloud control by using digital twin technology, and collects actual operation data and simulation operation data for analysis and calculation, which can effectively weaken the inconsistency of the current battery system and reduce the disturbance of the inconsistency of the future battery system, so as to manage the inconsistency of the power battery and control the development direction of the future inconsistency.

[0005] However, the existing battery design and management method has low precision, high cost and low sensitivity. SUMMARY

[0006] The present application aims to overcome the shortcomings of the prior art, and provides a power battery design and management method based on digital twin, which can reduce cost, realize agile development, reduce the workload and blindness of parameter identification, has high precision and stability, and can realize full life cycle closed loop control of power battery.

[0007] The application provides a power battery design and management method based on digital twinning, characterized in that the method comprises the following steps:

[0008] (1) a multi-physics field coupling mechanism model is established and mapped with a battery sample, and experimental data are used to calibrate the multi-physics field coupling mechanism model;

[0009] (2) the active subspace method is used to analyze the sensitivity of the calculation result data set of the multi-physics field mechanism model, and key parameters affecting the battery performance are screened out as main control parameters of the model;

[0010] (3) the main control parameters are taken as inputs, the simulation data set of the multi-physics field coupling mechanism model is taken as a training set, the simulation results and the neural network prediction results are used to determine a loss function, and a neural network efficient quantitative simulation model is established;

[0011] (4) the main performance parameters of the battery are quickly simulated based on the neural network efficient simulation model, part of the functions in the multi-physics field coupling mechanism model is combined with the neural network model, and a power battery twinning model capable of mapping physical entities is obtained.

[0012] Further, in the step (4), the part of the functions in the multi-physics field coupling mechanism model is at least one of side reaction calculation, heat production calculation and stress calculation.

[0013] Further, in the step (4), the input parameters of the power battery twinning model are actual operation data transmitted by a data interaction system and obtained through parameter identification.

[0014] Further, the method further comprises a step (6), specifically: based on the power battery twinning model, a secondary development of demand functions is performed on a digital twinning platform, and the results after function processing are fed back to realize full-life-cycle closed-loop management of the power battery.

[0015] Further, in the step (6), the demand functions include at least one of state estimation, capacity prediction and thermal runaway early warning.

[0016] Further, in the step (6), the demand functions further include main control parameter identification on actual operation data.

[0017] Further, the main control parameter identification method is at least one of a genetic algorithm, a differential evolution algorithm and a cuckoo algorithm.

[0018] Further, the multi-physics field mechanism model in the step (1) is coupling of at least two of an electrochemical reaction model, a thermal model, a mechanical stress model and a side reaction model.

[0019] The application also provides a digital twin power battery system, comprising: a power battery multi-physical field coupling mechanism model, a neural network efficient quantization simulation model, a digital twin cloud platform, a power battery physical entity, a master parameter identification module, and a power battery cloud function module.

[0020] The digital twin-based power battery design and management method can achieve:

[0021] (1) In the design stage of the power battery, a multi-physical field coupling mechanism model is established, the active subspace method is used to efficiently analyze the master control parameters and reaction mechanism of the model, and an efficient quantization simulation model based on a neural network is established for the active direction of the high-dimensional input parameter space. The model can realize rapid trial production of battery samples through parameter adjustment, and improve the design and optimization level of the power battery. In the evolution process of the whole life cycle of the power battery, a digital twin platform is established, and a power battery multi-physical field coupling digital model is established. Through the information interaction system and the physical space of the power battery entity, virtual mapping and data transmission are generated, which can realize accurate state estimation, evolution trend prediction, fault diagnosis and thermal runaway warning of the power battery, and can also provide experience feedback for the design of the power battery. The active subspace method of the multi-physical field coupling mechanism model simulation sensitivity analysis method only needs to calibrate the mechanism model with part of the experimental data to establish the neural network efficient simulation model, which greatly reduces the number of experiments and time cost in the design process of the power battery, and realizes agile development in the forward design process of the battery.

[0022] (2) The master control parameters obtained by the application can not only be used as the input of the twin model, but also reduce the workload and blindness of parameter identification. Moreover, because they have stronger representativeness, they can be used as storage variables to record the rules of the whole life cycle evolution process of the battery.

[0023] (3) The digital twin platform of the application can cooperate with the vehicle-end battery management system to realize efficient management of the power battery. The established twin model can be iterated with the whole life cycle evolution of the battery, and the accuracy will not diverge with time.

[0024] (4) The digital twin platform established by the application can store and calculate various parameters of the battery evolution process, and form certain rules and experience feedback to guide the battery design, and truly realize the whole life cycle closed-loop control of the power battery. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 The figure is a digital twin-based power battery design and management scheme;

[0026] Figure 2 The figure is a power battery multi-physical field coupling mechanism model scheme;

[0027] Figure 3A scheme diagram for a neural network efficient quantization simulation model scheme;

[0028] Figure 4 A scheme diagram for master control parameter identification using a differential evolution algorithm. DETAILED DESCRIPTION

[0029] The specific implementation of the present application is described in detail below, and it is necessary to point out here that the following implementation is only for further illustration of the present application and cannot be understood as a limitation on the protection scope of the present application. Some non-essential improvements and adjustments made by those skilled in the art to the present application according to the above content of the present application still belong to the protection scope of the present application.

[0030] The present application provides a power battery design and management method based on digital twinning, and the specific implementation mode is as shown in the accompanying Figures 1-4 The present application provides a power battery design and management method based on digital twinning, and the specific implementation mode is as shown in the accompanying Figure 1 The present application provides a power battery design and management method based on digital twinning, and the specific implementation mode is as shown in the accompanying Figure 2 The present application provides a power battery design and management method based on digital twinning, and the specific implementation mode is as shown in the accompanying Figure 3 The present application provides a power battery design and management method based on digital twinning, and the specific implementation mode is as shown in the accompanying Figure 4 The present application provides a power battery design and management method based on digital twinning, and the specific implementation mode is as shown in the accompanying

[0031] The present application provides a power battery design and management method based on digital twinning, which can combine the functions of the neural network efficient quantization simulation model and the multi-physical field coupling mechanism model established in the design stage, build a virtual model for the digital twinning platform, and realize efficient management of the power battery throughout its life cycle. The physical entity of the real world running power battery and the digital twinning model realize data transmission and iterative updating through a data interaction system, and the identified master control parameters can be stored and calculated, or used as feature parameters of the model for algorithm development. Since the twinning model has real-time performance, it not only can infinitely approach the physical entity but also can realize synchronous evolution, and through cloud function development and expansion, it can realize state estimation, life prediction, thermal runaway early warning and other functions.

[0032] Figure 1 The present application provides a power battery design and management method based on digital twinning, and the specific implementation mode is as shown in the accompanying

[0033] A multi-physical field coupling mechanism model mapped with a battery sample is established in the design stage, and the model is calibrated using experimental data, so that the power battery twinning model in the design stage can be obtained, which can accurately reflect the influence law of different physical and chemical parameters on the battery performance;

[0034] Because the model requires numerous parameters that are difficult to measure in their entirety, the active subspace method is used to perform sensitivity analysis on the computational results dataset of the multiphysics mechanism model, selecting key parameters affecting battery performance as the model's master parameters. The active subspace (AS) method is a dimensionality reduction method based on data mapping. It uses the partial covariance matrix eigenvalue decomposition of gradients to obtain active feature directions, constructing a low-dimensional subspace of the input parameters. Active variables are obtained by projecting the input parameters into this subspace, thus reducing the dimensionality of the input parameters. This method differs from other commonly used sensitivity dimensionality reduction methods in that sensitivity analysis retains the input parameters with the greatest impact, while subspace analysis retains the directions within the space of input parameters with the greatest impact. These directions are linear combinations of the parameters themselves, thus achieving maximum dimensionality reduction while maintaining accuracy. Furthermore, the component values ​​of the active direction vector provide global sensitivity information of the target quantity relative to the input parameters.

[0035] By using the master control parameters as input and the simulation dataset of the mechanistic model as the training set, a loss function is established using simulation results and neural network prediction results to build a highly efficient quantitative simulation model. This model can adjust the input of the master control parameters to obtain predicted values ​​of different battery performance parameters, enabling agile design and rapid optimization of power batteries.

[0036] The obtained neural network high-efficiency simulation model can quickly simulate the main performance parameters of the battery. By combining some functions of the multiphysics coupling mechanism model (such as side reaction calculation, heat generation calculation, and stress calculation) with the neural network model, a twin model that can map physical entities is obtained. The input parameters of this model are obtained by parameter identification from the actual operating data transmitted by the data interaction system.

[0037] Based on the digital twin model, secondary development of required functions can be carried out on the platform to realize functions such as state estimation, capacity prediction, and thermal runaway early warning. The cloud-based functional modules use the twin model as the domain knowledge model, co-evolves with the physical entity, and is continuously updated over time, resulting in more accurate and stable effects, and facilitating functional development.

[0038] The digital twin platform can store and calculate precise evolution of master control parameters, state performance estimates, and other features. Therefore, it can feed the results back to the battery design department, providing product design and optimization directions and quality control management basis, and truly realize closed-loop management of the entire life cycle of power batteries.

[0039] The active subspace method can be used in an m-dimensional input parameter space x∈R m The low-dimensional structure corresponding to the target quantity, i.e., the output quantity f(x), is identified, where the gradient of f with respect to x is maximized along the low-dimensional direction. By performing eigenvalue decomposition on the covariance matrix of the gradient of f(x), the active subspace of the input space can be obtained:

[0040]

[0041] In the formula, π is the joint probability density function (PDF) of the input parameters, C is a symmetric positive semi-definite matrix, and W = [w1,...,w m ] is the eigenvector matrix, Λ=diag(λ1,...,λ m Let be a diagonal matrix consisting of corresponding eigenvalues, where λ1,...,λ m Arranged in descending order of eigenvalue λ n Much greater than λ n+1 , that is, λ n >>λ n+1 Where n is less than m, meaning that f(x) has a large gradient in the first n eigenvector directions, while remaining almost unchanged in the last mn eigenvector directions. The low-dimensional space formed by the first n eigenvectors is the active subspace S. If the gradient of the mapping f(x) can be directly obtained in the simulation calculation, then the active subspace can be obtained by random sampling in the input space through the following steps:

[0042] 1) Based on the input parameter x∈R m The probability density distribution π(x) is obtained by sampling M samples {x} from the input space. (1) ,...,x (M)};

[0043] 2) Calculate the gradient for each sample point. In the formula, i = 1,...,M;

[0044] 3) Estimation matrix C:

[0045] 4) To Perform eigenvalue decomposition

[0046] 5) Divide the eigenma matrix and eigenvector matrix into blocks based on the magnitude of the eigenvalues:

[0047]

[0048] In the formula, and Includes the first n feature vectors Includes the remaining feature vectors;

[0049] 6) The space formed by expanding the eigenvectors is the active subspace, and the corresponding active vector is... The input parameter dimension is reduced from m-dimension to n-dimension.

[0050] Figure 2 The power battery multi-physical field coupling mechanism model scheme of the application is shown. The multi-physical field mechanism model is an electrochemical reaction model, a thermal model, a mechanical stress model, a side reaction model and coupling of different models. The application is described by taking an electrochemical thermal coupling model as an example:

[0051] 1) Solid phase lithium ion diffusion equation: the lithium ion concentration dimensionless balance equation is established based on the second Fick's law, and the lithium ion diffusion process in the active material particles in the positive and negative electrode regions is described;

[0052]

[0053] In the formula: and respectively represent the solid phase lithium ion concentration of the positive electrode region and the negative electrode region, is the solid phase diffusion coefficient of the positive and negative electrode regions, r is a size parameter along the radial direction of the spherical particle, and t is the time of the lithium ion diffusion process in the spherical particle.

[0054] 2) Liquid phase lithium ion diffusion equation: the lithium ion diffusion process in the electrolyte in the positive and negative electrode regions and the separator region is described;

[0055]

[0056]

[0057]

[0058] In the formula: D e is the effective diffusion coefficient of the liquid phase lithium ion, and are the electrolyte volume fractions of the positive and negative electrode regions, is the anion transfer number, is the negative electrode liquid phase lithium ion concentration, is the electrolyte liquid phase lithium ion concentration, is the positive electrode liquid phase lithium ion concentration, F is the Faraday constant, L - and L + are the positive and negative electrode thicknesses, and I(t) is the time-varying current.

[0059] 3) Solid phase potential equation: the distribution of the solid phase potential in the positive and negative electrode regions is described;

[0060]

[0061] In the formula: σ eff,± is the converted actual effective electrode conductivity, φs is the solid phase potential, Js is the current density at the electrode and current collector interface.

[0062] 4) Liquid phase potential equation: describes the distribution of liquid phase potential in the positive and negative electrode and separator regions;

[0063]

[0064] wherein: φl represents the liquid phase potential of the positive or negative electrode; κ f represents the equivalent conductivity of the liquid phase; R represents the ideal gas constant; and T represents the temperature; represents the conductivity of the liquid phase; L sep is the thickness of the electrolyte separator region; c e is the liquid phase lithium ion concentration. The first term on the right side of the equation represents the impedance potential caused by the liquid phase resistance of the electrolyte, and the second term represents the overpotential (concentration polarization voltage) caused by the lithium ion concentration gradient in the electrolyte.

[0065] 5) Butler-Volmer kinetics equation: describes the electrochemical reaction kinetics process at the solid-liquid phase interface inside the battery;

[0066]

[0067] wherein: α = a a or a c , a a and a c are the transfer coefficients of the anion and cation, respectively, is the electrode overpotential, a ± is the volume fraction of the positive and negative electrodes, is the exchange current density, which is expressed as follows:

[0068]

[0069] wherein: is the solid phase particle surface concentration, c e is the liquid phase electrolyte concentration, is the maximum solid phase particle concentration, k ± is the liquid phase lithium ion concentration, and the liquid phase conductivity can be calculated by fitting.

[0070] 6) Charge conservation equation: describes the charge balance phenomenon inside the battery system;

[0071]

[0072] wherein, is the exchange current density of the positive and negative electrodes.

[0073] 7) Temperature correction equation: describes the change rule of parameters such as conductivity and diffusion coefficient affected by temperature based on Arrhenius formula.

[0074]

[0075] In the formula, ψ ref is the property value at the reference temperature T ref . The temperature sensitivity of each single property is controlled by its corresponding activation energy .

[0076] Figure 3 The neural network efficient quantization simulation model scheme provided by the application is shown in the figure. The main control parameters are input, the simulation data set of the mechanism model is the training set, and the loss function is established by using the simulation results and the neural network prediction results. The neural network model can realize accurate fitting of high-dimensional nonlinear systems through the connection between neurons with weights, the opening and closing of the door, the restriction of the activation function, etc., and quickly solve by using error back propagation algorithm, which meets the demand of the application for calculation efficiency and accuracy.

[0077] In the digital twin platform, an important function is to identify the main control parameters of the actual operation data. The parameter identification technology is essentially an optimization solution to multiple target parameters. In the application, intelligent algorithms including but not limited to genetic algorithm, differential evolution algorithm, and cuckoo algorithm can be used for parameter identification. Figure 4 Taking the differential evolution algorithm as an example, the main control parameter identification scheme provided by the application is described as follows:

[0078] The differential evolution (DE) algorithm is derived from the early genetic algorithm (GA), which simulates crossover, mutation and reproduction in genetics to design genetic operators. The differential evolution algorithm is driven by mutation and selection processes. The mutation process includes mutation and crossover operations, which are designed to explore or exploit the search space, while the selection process is used to ensure that closer individual information can be further utilized.

[0079] In addition, the digital twin power battery system mainly includes: power battery multi-physical field coupling mechanism model (design stage model), neural network efficient quantization simulation model (use stage model, i.e. digital twin cloud virtual model), digital twin cloud platform, power battery physical entity, main control parameter identification module, power battery cloud function module, etc.

[0080] The power battery multi-physical field coupling mechanism model is a reference numerical model in the design stage, which includes an electrochemical reaction model for describing the complex nonlinearity inside the battery, a thermal model, a mechanical stress model, a side reaction model, and a coupling mechanism of different models. The classical electrochemical reaction model includes a pseudo two-dimensional model, a single particle model, a homogeneous model, etc. The thermal model includes a Fourier heat transfer equation, a Bonetti equation, etc. The mechanical stress model includes a particle crushing model, a stress model, etc. The side reaction model includes an SEI film generation model, a lithium precipitation model, etc. The solving algorithm of the model includes but is not limited to a finite difference method, a finite volume method, a finite element method, etc. Since the multi-physical field coupling mechanism model is established by applying physical and chemical equations such as porous electrode theory, concentrated solution theory, and heat production and heat transfer theory, the internal reaction process of the battery can be described in detail. However, the calculation includes a large number of nonlinear partial differential equations and strict microscopic parameters, initial conditions, and boundary conditions, so it is mainly used for battery design calculation and performance simulation.

[0081] The neural network efficient quantization simulation model relies on the simulation calculation results of the multi-physical field coupling mechanism model, uses the active subspace method to analyze the uncertainty of the high-dimensional input parameters of the mechanism model, obtains the transfer law and parameter sensitivity of the uncertainty of each part of the model parameters and the initial conditions and boundary conditions, and reveals the main control physical mechanism of the internal reaction. The high-sensitivity input parameters retained after screening are used as inputs, the main control parameters are used as inputs, the simulation data set of the mechanism model is used as a training set, a loss function is established by using the simulation results and the neural network prediction results, and the neural network model can realize the design target and efficient quantization simulation of the internal state of the battery.

[0082] The digital twin cloud platform is a node connected to the physical entity and the cloud twin model. On the one hand, the platform stores the multi-source historical data generated in the running process of the physical entity collected through sensors, system backends, etc. These data are stored in the cloud database of the digital twin platform and can be used for model training and standard setting. On the other hand, the cloud server is used to continuously calculate various parameters or indicators of various models to obtain information prediction, analysis, and evaluation at the current or future time period, realize the value realization and sharing of information, and achieve the value realization and sharing of information.

[0083] The power battery entity is one of the components of the digital twin platform, which is the actual running power battery in the physical space, and is used to obtain real-time parameters of the battery, including voltage, current, temperature, humidity, external stress, etc. When the digital twin system defines the function of the power battery, the power battery becomes a controlled object. The power battery entity includes: a battery pack, a battery management system, a thermal management system, a vehicle T-BOX, a CAN communication module, and a vehicle model.

[0084] The main control parameter identification module is one of the components of the digital twin platform, can identify the main control parameters of the twin model from the massive data uploaded by the physical entity, is an important module for realizing the mapping of the physical entity and the twin model, and can also be used as the feature parameters of the power battery to send to the power battery cloud function module.

[0085] The power battery twin model is one of the components of the digital twin platform, is the synchronous digital mapping of the physical entity in the twin platform, and is also the basis of the cloud function module. The twin model of the present application mainly refers to the trained neural network efficient simulation model, and part of the function of the multi-physical field coupling mechanism model as a supplement.

[0086] The power battery cloud function module is one of the components of the digital twin platform, mainly composed of some micro-service functions for the user end, such as state estimation, life prediction, thermal runaway prediction, charging control, etc. In order to ensure the efficient management of the power battery system, the design of this part needs to meet the definition of functional safety and information safety.

[0087] Although the exemplary embodiments of the present application have been described for the purpose of illustration, those skilled in the art will understand that various modifications, additions and substitutions can be made without departing from the scope and spirit of the application disclosed in the appended claims, and all such changes shall fall within the scope of the appended claims, and the product claimed by the present application and the steps in the method claimed by the present application can be combined in any form. Therefore, the description of the embodiments disclosed in the present application is not intended to limit the scope of the present application, but to describe the present application. Accordingly, the scope of the present application is not limited by the above embodiments, but is defined by the claims or their equivalents.

Claims

1. A digital-twin-based power battery design and management method, characterized in that, Comprise the following steps: (1) Establish a multi-physics field coupling mechanism model mapped with the battery sample, and calibrate the multi-physics field coupling mechanism model using experimental data; (2) Use the active subspace method to analyze the sensitivity of the calculation result data set of the multi-physics field mechanism model, and select the key parameters affecting the battery performance as the main control parameters of the model; (3) Take the main control parameters as input, and take the simulation data set of the multi-physics field coupling mechanism model as the training set, use the simulation results and neural network prediction results to determine the loss function, and establish a neural network high-efficiency quantitative simulation model; (4) Based on the neural network high-efficiency simulation model, the performance parameters of the battery are simulated quickly, part of the functions in the multi-physics field coupling mechanism model are combined with the neural network model, and a power battery twin model capable of mapping physical entities is obtained; The part of the function in the multi-physics field coupling mechanism model in step (4) is at least one of side reaction calculation, heat production calculation and stress calculation; The input parameters of the power battery twin model in step (4) are obtained by parameter identification from the actual running data transmitted by the data interaction system, and the input parameters include voltage, current, temperature, humidity and external stress; The multi-physics field mechanism model in step (1) is at least two or more of the coupling of electrochemical reaction model, thermal model, mechanical stress model and side reaction model.

2. The method of claim 1, wherein: Further comprising step (5), specifically: based on the power battery twin model, the secondary development of demand function is carried out on the digital twin platform, and the results after function processing are fed back to realize the closed-loop management of the whole life cycle of the power battery.

3. The method of claim 2, wherein: The demand function in step (5) includes at least one of state estimation, capacity prediction and thermal runaway early warning.

4. The method of claim 3, wherein: The demand function in step (5) further comprises main control parameter identification on actual running data.

5. The method of claim 4, wherein: The method of main control parameter identification adopts at least one of genetic algorithm, differential evolution algorithm and cuckoo algorithm.

6. A digital twin power battery system implemented by the digital twin-based power battery design and management method of any one of claims 1-5. Comprise: A multi-physics field coupling mechanism model of a power battery, a neural network high-efficiency quantitative simulation model, a digital twin cloud platform, a power battery physical entity, a main control parameter identification module, and a power battery cloud function module.

Images