Power battery performance degradation prediction method and system based on electrochemical-thermal-mechanical-neural network model

Abstract

The invention relates to a power battery performance degradation prediction method and system based on an electrochemical-thermal-mechanical-neural network model, and belongs to the technical field of battery management. According to the method, an enhanced single particle model considering liquid phase diffusion, an integrated thermal model, a particle mechanical stress model caused by concentration gradient and an aging mechanism model considering SEI growth, lithium precipitation and active material loss are coupled to establish ETMD, and based on the ETMD, the identification efficiency of aging model parameters is low aiming at the problem that a PSO algorithm is low. Under a circulating aging experiment of a limited path, the characteristic parameters of the aging model are obtained by using the feedforward neural network, and the parameters are substituted into the ETMD model again to predict the capacity, the power attenuation curve and the corresponding main attenuation mechanism under the current path, so that the prediction precision and the interpretability of the performance attenuation of the battery under the full path are improved. According to the method, the modeling capability of battery performance degradation under different working conditions can be effectively improved, and a theoretical basis and technical support are provided for residual life prediction and safety management of the power battery.

Description

Technical Field

[0001] This invention belongs to the field of battery management technology and relates to a method and system for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model. Background Technology

[0002] In electric and hybrid vehicles, accurate prediction of battery performance is crucial for ensuring system safety and improving efficiency. During service, battery capacity and power gradually decline. Experimental data shows that under different aging paths (e.g., temperature, rate stress conditions), batteries may exhibit significant differences in power retention and remaining lifespan when degrading to the same State of Health (SOH). This phenomenon of "same SOH, different performance" places higher demands on battery management systems and increases the uncertainty of battery applications. It can easily lead to the vehicle system failing to achieve the expected power output or charging performance, thus affecting overall vehicle reliability and user experience.

[0003] Currently, methods for modeling battery performance degradation mainly include: statistical methods based on empirical formulas, mechanistic models, and data-driven methods. Empirical models can achieve rapid calculations; however, empirical models based on simple formula fitting of experimental data cannot predict the differences in battery performance degradation under different aging paths. Mechanistic models can describe the battery aging process from the perspective of side reaction mechanisms; however, they contain a large number of electrochemical parameters, which are complex to identify and computationally intensive, resulting in long processing times. In addition, some studies have attempted to introduce machine learning methods for degradation modeling, such as feedforward neural networks, convolutional neural networks, and recurrent neural networks. However, due to the lack of consideration for the coupling mechanisms under multi-path aging, their generalization ability is insufficient, making it difficult to accurately predict the performance change trends under various stress conditions.

[0004] Therefore, there is an urgent need to develop a prediction method that integrates mechanism modeling and data-driven approaches to characterize the impact mechanism of different aging paths on the power degradation and remaining lifespan of power batteries. This is to address the problem that power batteries aged to the same state of equilibrium (SOH) at different temperatures and rates have different power degradation and remaining lifespans, which leads to the inability of actual battery applications to reach the preset level. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method and system for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model. This method and system construct a multiphysics coupled model and introduce a neural network to learn the dependence of electrochemical aging parameters on the aging path, balancing prediction accuracy and computational efficiency. This effectively improves the ability to model battery performance degradation under different operating conditions, providing a theoretical basis and technical support for predicting the remaining life and safe management of power batteries.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model, the method specifically includes the following steps:

[0008] S1: Select the power battery to be tested, collect and organize relevant technical parameters, establish an electrochemical-thermal-mechanical-degradation (ETMD) model for this type of power battery, and determine the model input and output parameters;

[0009] S2: Perform cycle aging test and calendar aging test on the battery. In the cycle aging test, a performance reference test (PRT) is performed on the power battery under test every 10 cycles (one month of storage in the calendar aging test), and battery status data is collected, including terminal voltage, current, temperature, resistance and capacity data.

[0010] S3: Based on the experimental voltage and temperature dataset collected from the fresh battery RPT, perform parameter identification on the electrochemical-thermal model, determine the characteristic parameters of the fresh battery electrochemical-thermal model under different operating conditions, and calculate the terminal voltage V based on the electrochemical-thermal model. t Estimate the relationship between the battery average temperature T c Estimate and verify with experimental data;

[0011] S4: Based on the battery capacity, resistance, and temperature database under cycle and calendar aging tests, the parameters of the aging mechanism model are identified to obtain the characteristic parameters of different aging mechanism models under different paths. Based on the aging model, the capacity C and resistance R are estimated.

[0012] S5: Select 80% of the experimental aging paths to train a neural network from the aging path to the characteristic parameters of the aging mechanism model, establish an electrochemical-thermal-mechanical-aging-neural network (ETMD-Neural Network, ETMD-NN) model, obtain the characteristic parameters of the aging model under different aging paths based on the NN model, substitute them into the ETMD model to estimate the growth path of battery capacity C and resistance R, and compare the capacity trajectory and power decay trajectory estimated by the ETMD-NN model under the remaining experimental paths used for verification with the experimental results;

[0013] S6: Based on the established ETMD-NN model, predict the evolution trajectory of battery state of health (SOH), aging state of resistance (ASOR), aging state of peak power (ASOP), and aging state of average temperature (ASOT) at the end of operation under different aging paths, and determine the main aging mode under the current path.

[0014] Furthermore, in step S1, the electrochemical sub-model used in the power battery is the extended single-particle (SP) model, which uses a single particle to represent the behavior of all particles and considers the non-uniform diffusion of lithium ions in the electrolyte to establish a simplified electrochemical model; the thermal model of the power battery is a one-dimensional (1-D) unsteady-state heat generation and transfer model; the mechanical model is the calculation of particle stress caused by the lithium ion concentration gradient; the aging model considers the growth of the solid electrolyte interface (SEI) film, lithium plating (LP), and loss of active material (LAM).

[0015] Furthermore, step S2 specifically includes:

[0016] S21: Set the test temperature and let it stand until thermal equilibrium is reached;

[0017] S22: Based on the required charging and discharging paths for the test, three batteries are used for each path, with upper and lower cutoff voltages ranging from 4.2V to 2.75V. The cycling paths consider different ambient temperatures (-20℃, -5℃, 10℃, 25℃, 40℃), charge / discharge rates (0.2C charge-0.2C discharge; 1.5C charge-1.5C discharge; 2C charge-2C discharge; 0.2C charge-1.5C discharge), and depth of charge / discharge (DOD), such as 0% DOD-100% DOD, 0% DOD-50% DOD, for grouped battery cycling tests. During calendar aging, the batteries are stored at -5℃, 25℃, and 40℃ respectively.

[0018] S23: Perform a set of RPTs on the battery every 10 cycles / every month of storage; including capacity testing at standard charge / discharge current at 25°C; 1 / 20C trickle charge / discharge testing at 25°C; Hybrid Pulse Power Characteristic (HPPC) testing at 25°C and 1C pulse current for different SOCs; charge / discharge temperature rise testing at 25°C and 1C current; and battery thickness testing.

[0019] S24: Summarize and process the acquired experimental data to form data that can be used for later neural network training and validation of the ETMD-NN model.

[0020] Furthermore, in step S3, the Particle Swarm Optimization (PSO) algorithm is used to set the output term as the average battery temperature T. c and terminal voltage V t The loss function is defined as the sum of four parts: first, the sum of the root mean square error (RMSE) of the simulated voltage and the measured voltage under different operating conditions; second, the sum of the RMSE of the simulated state of charge (SOC) and the measured state of charge (SOC) of the positive electrode under different operating conditions; third, the sum of the RMSE of the simulated state of charge (SOC) and the measured state of charge (SOC) of the negative electrode under different operating conditions; and fourth, the sum of the RMSE of the simulated average temperature (Tc) of the battery and the measured average temperature of the battery under different operating conditions. This is used to identify highly sensitive parameters in the electrochemical-thermal model.

[0021] Furthermore, in step S4, the PSO algorithm is used to set the output term as open circuit voltage (OCV), and the loss function is defined as the RMSE of the simulated OCV and the 1 / 20C trickle charge / discharge terminal voltage in the RPT test, to identify the electrode health status (eSOH, eSOH) parameters.

[0022] Furthermore, step S5 specifically includes:

[0023] S51: Determine the aging path characteristics, including charge / discharge current I, ambient temperature T, and charge / discharge depth DOD, as inputs to the NN. The characteristic parameters of the aging model under this path are the target values.

[0024] S52: Under the validation set path, compare the accuracy of the aging model parameters output by the NN model with the aging model parameters obtained by parameter identification using the PSO method.

[0025] The present invention also provides a power battery performance degradation prediction system based on an electrochemical-thermal-mechanical-neural network model.

[0026] The beneficial effects of this invention are as follows:

[0027] (1) The established ETMD model can achieve joint characterization of the macroscopic aging performance and microscopic mechanism of the battery. On a macroscopic level, the multi-physics model framework can output SOH, ASOR, ASOP and ASOT performance indicators; on a microscopic level, the dominant mechanism under different aging paths can be determined based on the eSOH method.

[0028] (2) A multi-step, multi-objective aging mechanism model parameter identification framework based on eSOH, resistance, and expansion is proposed, which can achieve high-precision aging parameter identification and is applicable to different types of batteries.

[0029] (3) A physical sensing neural network framework suitable for different aging paths is proposed to realize the fast mapping between aging paths and aging model parameters, thereby improving the efficiency of aging model parameter identification and the accuracy of battery performance degradation prediction under the whole path.

[0030] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0031] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0032] Figure 1 This is a flowchart of the method of the present invention;

[0033] Figure 2 This is a simplified diagram of the ETMD-NN model according to an embodiment of the present invention;

[0034] Figure 3This is a flowchart of the experimental data acquisition process in step S2 of an embodiment of the present invention;

[0035] Figure 4 This is a flowchart of the multi-step multi-target parameter identification method according to an embodiment of the present invention;

[0036] Figure 5 This is a flowchart of step S5 in an embodiment of the present invention. Detailed Implementation

[0037] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0038] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0039] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not 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, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0040] This invention couples an enhanced single-particle model considering liquid-phase diffusion with an integrated thermal model, a particle mechanical stress model caused by concentration gradient, and an aging mechanism model considering SEI growth, lithium plating, and active material loss to establish an ETMD. Based on this, to address the problem of low efficiency in identifying aging model parameters by the PSO algorithm, a feedforward neural network is used to obtain the characteristic parameters of the aging model under finite-path cyclic aging experiments. These parameters are then re-substituted into the ETMD model to predict the capacity and power decay curves under the current path, as well as the corresponding main decay mechanisms, thereby improving the accuracy and interpretability of battery performance decay prediction under the entire path. Figure 1 The flowchart of the method of the present invention is shown in the figure. The method for predicting the performance degradation of a power battery based on an electrochemical-thermal-mechanical-neural network model of the present invention includes the following steps:

[0041] S1: Select the power battery to be tested, collect and organize relevant technical parameters, establish an electrochemical-thermal-mechanical-degradation (ETMD) model for this type of power battery, and determine the model input and output parameters;

[0042] S2: Perform cycle aging test and calendar aging test on the battery. In the cycle aging test, a performance reference test (PRT) is performed on the power battery under test every 10 cycles (one month of storage in the calendar aging test), and battery status data is collected, including terminal voltage, current, temperature, resistance and capacity data.

[0043] S3: Based on the experimental voltage and temperature dataset collected from the fresh battery RPT, perform parameter identification on the electrochemical-thermal model, determine the characteristic parameters of the fresh battery electrochemical-thermal model under different operating conditions, and calculate the terminal voltage V based on the electrochemical-thermal model. t Estimate the relationship between the battery average temperature T c Estimate and verify with experimental data;

[0044] S4: Based on the battery capacity, resistance, and temperature database under cycle and calendar aging tests, the parameters of the aging mechanism model are identified to obtain the characteristic parameters of different aging mechanism models under different paths. Based on the aging model, the capacity C and resistance R are estimated.

[0045] S5: Select 80% of the experimental aging paths to train a neural network from the aging path to the characteristic parameters of the aging mechanism model, establish an electrochemical-thermal-mechanical-aging-neural network (ETMD-Neural Network, ETMD-NN) model, obtain the characteristic parameters of the aging model under different aging paths based on the NN model, substitute them into the ETMD model to estimate the growth path of battery capacity C and resistance R, and compare the capacity trajectory and power decay trajectory estimated by the ETMD-NN model under the remaining experimental paths used for verification with the experimental results;

[0046] S6: Based on the established ETMD-NN model, predict the evolution trajectory of battery state of health (SOH), aging state of resistance (ASOR), aging state of peak power (ASOP), and aging state of average temperature (ASOT) at the end of operation under different aging paths, and determine the main aging mode under the current path.

[0047] In this embodiment:

[0048] S1: Select the power battery to be tested and establish a continuous electrochemical-thermal-mechanical-aging model of the power battery in the time domain. The model coupling mechanism is described in [link to relevant documentation]. Figure 2 And determine the parameters of the dynamic model. Specifically...

[0049] Physical meaning and governing equations of the single-particle model considering liquid-phase diffusion:

[0050] (1) Solid phase lithium ion distribution

[0051] The diffusion of lithium ions in solid spherical particles follows Fick's second law:

[0052]

[0053] Its initial conditions are:

[0054] c s,i (r,t)| t＝0 =c s,0

[0055] Its boundary conditions are:

[0056]

[0057] In ESPM, the orifice wall flux in the x-direction of the electrode thickness is replaced by the average value:

[0058]

[0059] (2) Liquid phase lithium ion concentration distribution

[0060] The liquid phase concentration distribution of lithium ions in the positive and negative electrode regions conforms to:

[0061]

[0062] The liquid phase concentration distribution in the diaphragm region conforms to:

[0063]

[0064] The boundary conditions are:

[0065]

[0066]

[0067] (3) Electrode Dynamics

[0068] In lithium-ion batteries, the insertion / extraction of lithium ions is a charge transfer reaction occurring at the solid / liquid interface. This charge transfer reaction can be described based on the Butler-Volmer kinetic equations:

[0069]

[0070] i0 is the current exchange density related to the concentration gradients of the solid and liquid phases, which can be expressed as:

[0071]

[0072] Terminal voltage calculation:

[0073]

[0074] Physical significance and governing equations of the thermal model:

[0075] The heat sources inside a battery mainly consist of three parts: ohmic heat, reaction heat, and polarization heat.

[0076] q gen =q ohm +q rxn +q pol

[0077]

[0078] Physical significance and governing equations of the mechanical model:

[0079] Assuming the particles are isotropic and linearly elastic materials, the resulting stresses are independent. The radial and tangential stress-strain components are described as follows:

[0080]

[0081] In the formula, ξ r ξ θ Radial strain and tangential strain, respectively, σ r σ θ These are radial stress and tangential stress, respectively. Due to the symmetry of spherical particles, the ξ on the surface of small deformable particles... r and ξ θ Each by and It is determined that μ is the radial displacement. The diffusion rate of lithium ions in the particles is much slower than the elastic deformation rate, therefore the elastic deformation can be considered as a static mechanical equilibrium.

[0082]

[0083] The boundary conditions for the particle center and the surface are:

[0084]

[0085] Therefore, the radial and tangential stresses caused by diffusion are strongly related to the lithium-ion concentration:

[0086]

[0087] Therefore, the total strain is expressed as

[0088]

[0089] The change in particle radius after lithium insertion or extraction can be determined by defining the equation r = R. p To estimate:

[0090]

[0091] Considering that the thickness of the anode and cathode layers is much smaller than the length and width of the battery, the thickness variation is estimated by the volume variation as follows:

[0092]

[0093] Among them, L p is the thickness of the layer fraction represented by the particles, and n is the number of anode and cathode layers in the pouch cell.

[0094] Physical significance and governing equations of the aging model:

[0095] (1) SEI growth

[0096]

[0097] η SEI =η n +Un (c ss )-U SEI

[0098]

[0099] SEI growth thickness increase:

[0100]

[0101] (2) Lithium plating

[0102]

[0103] i 0,pl =k pl c e (c ss -c s,avg ) / c s,max F

[0104] η pl =η n +U n (c ss )

[0105]

[0106] Increase in lithium plating layer thickness:

[0107]

[0108] (3) Loss of active materials

[0109] The main source of active material loss is the diffusion-induced stress caused by the large lithium-ion concentration gradient inside the graphite particles, accompanied by the loss of active material volume fraction ε. s This decreases, thus affecting the capacity of both the positive and negative electrodes.

[0110]

[0111] For a single particle of the positive or negative electrode, the volume fraction ε of the active material s The rate of change is:

[0112]

[0113] (4) Internal resistance growth

[0114]

[0115] (5) Loss of lithium inventory

[0116] I side =AL n (j SEI +jpl )

[0117]

[0118] (6) Loss of active materials

[0119]

[0120] (7) Thickness increase

[0121] δ irrev =b SEI δ SEI +b pl δ pl +b T (TT ref )+b L dL th .

[0122] See Figure 3 Step S2 specifically includes steps S21 to S24.

[0123] S21: Set the test temperature and let it stand until thermal equilibrium is reached;

[0124] S22: Charge and discharge according to the required test path, using 3 batteries for each path, with upper and lower cutoff voltages of 4.2V-2.75V;

[0125] The cycling path considers different ambient temperatures (-20℃, -5℃, 10℃, 25℃, 40℃), charge / discharge rates (0.2C charge-0.2C discharge; 1.5C charge-1.5C discharge; 2C charge-2C discharge; 0.2C charge-1.5C discharge), and depth of charge / discharge (DOD), such as 0% DOD-100% DOD, 0% DOD-50% DOD, for battery group cycling tests. During calendar aging, the batteries are stored at -5℃, 25℃, and 40℃ respectively.

[0126] S23: Perform a set of RPTs on the battery every 10 cycles / every month of storage;

[0127] Includes capacity testing at standard charge / discharge current at 25℃; 1 / 20C trickle charge / discharge testing at 25℃; Hybrid Pulse Power Characteristic (HPPC) testing at 25℃ and 1C pulse current for different SOCs; charge / discharge temperature rise testing at 25℃ and 1C current; and battery thickness testing.

[0128] S24: Summarize and process the acquired experimental data to form data that can be used for later neural network training and validation of the ETMD-NN model.

[0129] The specific loss function in step S3 is as follows:

[0130]

[0131] in M represents the set of parameters to be identified, and N represents the number of operating conditions used for parameter identification. i The data length representing the i-th working condition These represent the model simulation values ​​of voltage, positive SOC, and negative SOC under the i-th operating condition, respectively.

[0132] The specific method for identifying aging model parameters using eSOH in step S4 is as follows:

[0133] (1) Identification of eSOH parameters based on open-circuit voltage (OCV) model

[0134] The open-circuit potential (OCV) of a battery is the difference between the open-circuit potentials of its two electrodes, and can be expressed as:

[0135] V oc (t)=U p (y(t))-U n (x(y))

[0136] Where Up represents the open-circuit potential of the positive electrode, Un represents the open-circuit potential of the negative electrode, y represents the lithium intercalation state of the positive electrode material, and x represents the lithium intercalation state of the negative electrode material. The range of the lithium intercalation state is [0,1], and y is used here. 100 and x 100 The lithium intercalation state represents the battery's fully charged state, while y0 and x0 represent the lithium intercalation state's fully discharged state.

[0137] V min =U p (y0)-U n (x0)

[0138] V max =U p (y 100 )-U n (x 100 )

[0139] The relationship between battery SOC and lithium intercalation state can be expressed as:

[0140]

[0141] Where Q represents the amount of electricity (Ah) that the battery can release at this time, and C represents the available capacity of the battery.

[0142] The relationship between the capacities of the positive and negative electrodes (Cp and Cn) and the usable capacity of the battery can be expressed as follows:

[0143] C = C n (x 100 -x0)=C n (y 100 -y0)

[0144] The lithium intercalation state of the two electrodes can be written as a function of Q and the electrode parameters Θ = [Cp, Cn, y0, x0]:

[0145]

[0146] The final OCV model can be written as:

[0147]

[0148] Total lithium inventory n Li It can be represented as:

[0149]

[0150] The OCV model has four parameters Θ = [Cp, Cn, y0, x0], which are combined with the calculated lithium-ion inventory n. Li The total battery capacity C is named the eSOH parameter. These parameters change as the battery ages, so the OCV in the dataset can be used to identify the four parameters. Here, the PSO algorithm is still used for parameter identification, and the optimization objective function is the root mean square error between the measured OCV and the simulated OCV:

[0151]

[0152] Furthermore, in order to improve the accuracy of parameter identification, the optimization objective must also meet the following constraints:

[0153] V max =U p (y 100 )-U n (x 100 )

[0154] (2) Parameter identification of aging mechanism model based on eSOH parameter

[0155] 1) Define aging mechanism state parameters

[0156] The battery degradation state X is determined based on the battery aging mechanism, which includes the lithium-ion inventory n. Li SEI film thickness δ SEI Thickness δ of lithium plating layer PL and the volume fraction of positive and negative electrode active materials

[0157] 2) Define the aging mechanism parameter P to adjust for specific aging mechanisms. P contains different parameters under different aging mechanisms.

[0158] P = [p] 1 ,p 2 ,…,p n ]

[0159] 3) Define the parameters that need to be fitted based on the model output, namely the parameters such as eSOH, expansion amount and resistance extracted from OCV, expansion data and resistance data.

[0160] Y = [y 1 ,y 2 ,…,y n ]

[0161] 4) Run the particle swarm optimization (PSO) algorithm to estimate the parameters of the aging mechanism model.

[0162] Using the eSOH parameter identification method based on the OCV model described in S3(1), the SEI model was first adjusted using calendar aging data at multiple temperatures. In this case, the contributions of active material loss and lithium plating to aging are negligible. Then, multiple cyclic aging datasets with different current rates and discharge depths were used to simultaneously adjust the lithium plating and mechanical models.

[0163] The specific process is as follows: Figure 4 As shown:

[0164] 1) SEI Model Parameter Identification: Calendar-aged cell data from low-temperature (-5℃) and high-temperature (45℃) conditions were used to adjust for SEI growth-induced aging in order to analyze the temperature dependence of SEI growth. Therefore, the Arrhenius equation was used to further describe the temperature-dependent SEI growth reaction rate k. SEI In the aging mechanism of SEI growth, the aging mechanism parameter P needs to be identified. cal Including the initial reaction rate constant k of SEI 0,SEI SEI layer diffusion coefficient D SeI SEI reaction activation energy E a,SEI .

[0165] P cal =[k 0,SEI D SEI E a,SEI ]

[0166] Y contains the battery capacity C and the number of recyclable lithium batteries n. Li and the stoichiometric ratios y0 and x 100 .

[0167] Y cal =[C,nLi ,y0,x 100 ]

[0168] 2) Mechanical model parameter identification: Adjusting the active material loss model parameter β derived from mechanical stress. LAM,1 ,β LAM,2 ,m LAM, To accommodate battery capacity C and the amount of cyclic lithium n Li and electrode capacity C n and C p Therefore, P and Y are respectively:

[0169] P cyc =[β LAM,1,p ,β LAM,2,p β LAM,1,n ,β LAM,2,n ,m LAM ]

[0170] Y cyc =[C,n Li C n C p ]

[0171] 3) Parameter identification for the resistance growth model: The resistance increase of the battery during aging originates from the increased resistance caused by the thickness of the SEI film and lithium deposition. Therefore, the parameters that need to be identified include the ionic conductivity k of the SEI film and the lithium deposition film. SEI ,k pl Y represents the increase in battery resistance, which can be obtained using HPPC testing.

[0172] P resistance =[k SEI ,k pl ]

[0173] Y resistance =[dR]

[0174] 4) Identification of expansion model parameters: Irreversible expansion comes from the volume change caused by the thickness of SEI, the thickness of lithium plating, and the loss of battery active materials. Therefore, the parameters that need to be identified include coefficients from each thickness before the increase. Y is the battery test expansion amount.

[0175] P expansion =[b SEI ,b pl ,b r,p ,b r,n ]

[0176] Y expansion =[dδ]

[0177] See Figure 5 Step S5 specifically includes steps S51 to S52.

[0178] S51: Determine the aging path characteristics (charge / discharge current I, ambient temperature T, charge / discharge depth DOD) as input to the NN, and the target values ​​are the characteristic parameters of the aging model under this path.

[0179] S52: Under the validation set path, compare the accuracy of the aging model parameters output by the NN model with the aging model parameters obtained by parameter identification using the PSO method.

[0180] The battery health status (SOH), resistance aging status (ASOR), average battery temperature aging status at the end of the operation (ASOT), and peak power aging status (ASOP) in step S6 are defined as follows:

[0181]

[0182] Taking discharge as an example, battery power is limited by either voltage or rated current. The aging power (voltage protection limit) is determined by the voltage limitation:

[0183] P 1,age,dis =(U OCV,age -I volt,dis ×P age )×I volt,dis

[0184] in

[0185]

[0186] Aging power (current safety limit) determined by rated current limitation.

[0187] P 2,age,dis =(U OCV,age -I max,dis ×R age )×I max,dis

[0188] Among them I max,dis The maximum continuous current given in the battery specifications

[0189] Therefore, the actual maximum discharge power is

[0190] P age,max,dis =min(P 1,age,dis ,P 2,age,dis ).

[0191] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model, characterized in that, The method specifically includes the following steps: S1: Select the power battery to be tested, collect and organize relevant technical parameters, establish an electrochemical-thermal-mechanical-degradation (ETMD) model for this type of power battery, and determine the model input and output parameters; S2: Perform cycle aging test and calendar aging test on the battery. In the cycle aging test, a performance reference test (PRT) is performed on the power battery under test every 10 cycles to collect battery status data, including terminal voltage, current, temperature, resistance and capacity data. S3: Based on the experimental voltage and temperature dataset collected from the fresh battery RPT, perform parameter identification on the electrochemical-thermal model, determine the characteristic parameters of the fresh battery electrochemical-thermal model under different operating conditions, and calculate the terminal voltage V based on the electrochemical-thermal model. t Estimate the relationship between the battery average temperature T c Estimate and verify with experimental data; S4: Based on the battery capacity, resistance, and temperature database under cycle and calendar aging tests, the parameters of the aging mechanism model are identified to obtain the characteristic parameters of different aging mechanism models under different paths. Based on the aging model, the capacity C and resistance R are estimated. S5: Select 80% of the experimental aging paths to train a neural network from the aging path to the characteristic parameters of the aging mechanism model, establish an electrochemical-thermal-mechanical-aging-neural network (ETMD-Neural Network, ETMD-NN) model, obtain the characteristic parameters of the aging model under different aging paths based on the NN model, substitute them into the ETMD model to estimate the growth path of battery capacity C and resistance R, and compare the capacity trajectory and power decay trajectory estimated by the ETMD-NN model under the remaining experimental paths used for verification with the experimental results; S6: Based on the established ETMD-NN model, predict the evolution trajectory of battery state of health (SOH), aging state of resistance (ASOR), aging state of peak power (ASOP), and aging state of average temperature (ASOT) at the end of operation under different aging paths, and determine the main aging mode under the current path.

2. The method for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model according to claim 1, characterized in that, In step S1, the electrochemical sub-model used in the power battery is the extended single-particle (SP) model, which uses a single particle to represent the behavior of all particles and considers the non-uniform diffusion of lithium ions in the electrolyte to establish a simplified electrochemical model; the thermal model of the power battery is a one-dimensional (1-D) unsteady-state heat generation and transfer model; the mechanical model is the calculation of particle stress caused by the lithium ion concentration gradient; the aging model considers the growth of the solid electrolyte interface (SEI) film, lithium plating (LP), and loss of active material (LAM).

3. The method for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model according to claim 2, characterized in that, Step S2 specifically includes: S21: Set the test temperature and let it stand until thermal equilibrium is reached; S22: Charge and discharge according to the required test path, using 3 batteries for each path, with upper and lower cutoff voltages of 4.2V-2.75V; S23: Perform a set of RPTs on the battery every 10 cycles / every month of storage; S24: Summarize and process the acquired experimental data to form data that can be used for later neural network training and validation of the ETMD-NN model.

4. The method for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model according to claim 3, characterized in that, In step S3, the Particle Swarm Optimization (PSO) algorithm is used, and the output term is set to the average battery temperature T. c and terminal voltage V t The loss function is defined as the sum of four parts: first, the sum of the root mean square error (RMSE) of the simulated voltage and the measured voltage under different operating conditions; second, the sum of the RMSE of the simulated SOC (State of Charge) and the measured SOC of the positive electrode under different operating conditions; and third, the sum of the RMSE of the simulated SOC and the measured SOC of the negative electrode under different operating conditions. The parameters to be identified are:

4. The sum of the RMSE of the simulated average temperature Tc of the battery under different operating conditions and the measured average temperature of the battery, and the highly sensitive parameters in the electrochemical-thermal model are identified.

5. The method for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model according to claim 4, characterized in that, In step S4, the PSO algorithm is used to set the output term as open circuit voltage (OCV), and the loss function is defined as the RMSE of the simulated OCV and the 1 / 20C trickle charge / discharge terminal voltage in the RPT test. The electrode health status (eSOH, eSOH) parameters are then identified.

6. The method for predicting the performance degradation of power batteries based on an electrochemical-thermal-mechanical-neural network model according to claim 5, characterized in that, Step S5 specifically involves: S51: Determine the aging path characteristics, including charge / discharge current I, ambient temperature T, and charge / discharge depth DOD, as inputs to the NN. The characteristic parameters of the aging model under this path are the target values. S52: Under the validation set path, compare the accuracy of the aging model parameters output by the NN model with the aging model parameters obtained by parameter identification using the PSO method.

7. A power battery performance degradation prediction system based on an electrochemical-thermal-mechanical-neural network model, characterized in that, The system employs the method described in any one of claims 1 to 6.

Images