A lithium battery internal aging mode prediction method, system, device and storage medium

Abstract

This invention discloses a method, system, device, and storage medium for predicting the internal aging mode of a lithium battery, relating to the field of battery health status prediction technology. The method includes: acquiring external signals of the lithium battery; the external signals include the battery's voltage, current, temperature, and charging time; constructing a lithium battery aging mode prediction model; the lithium battery aging mode prediction model includes a lithium-ion battery aging mechanism module and a lithium-ion battery data-driven module; the lithium-ion battery data-driven module includes a two-layer LSTM neural network, a fully connected layer, and an attention mechanism layer; inputting the external signals of the lithium battery into the lithium battery aging mode prediction model to predict the internal aging mode of the lithium battery; enabling the battery aging mode recognition system to accurately predict the internal aging mode of the lithium-ion battery based on the external signals during lithium-ion battery operation.

Description

Technical Field

[0001] This invention relates to the field of battery health status prediction technology, specifically to a method, system, device, and storage medium for predicting the internal aging mode of a lithium battery. Background Technology

[0002] As the number of electric vehicles on the road continues to increase, consumers are paying more attention to the driving range and charging time of electric vehicles. This necessitates continuous improvements in the energy density and maximum charge / discharge rate of lithium-ion batteries. To meet the range requirements, battery electrodes are designed to be thicker and denser to accommodate more lithium ions. However, increasing the density and thickness of the negative electrode makes it more difficult for the electrolyte to fully penetrate the electrode, slowing down the diffusion of lithium ions within the electrode active material particles and leading to a decrease in the battery's current rate performance. Therefore, due to limitations in materials technology and processing technology, it is difficult to achieve both high energy density and high power density in electric vehicle batteries. Furthermore, under harsh operating conditions that differ significantly from their designed usage scenarios, lithium-ion batteries are prone to aging, resulting in a rapid decline in their capacity.

[0003] Currently, traditional aging mechanism models for lithium-ion batteries mostly employ pseudo-two-dimensional (P2D) electrochemical models. However, these models require the identification of numerous parameters, leading to a significant amplification of errors. Many parameters in P2D electrochemical models cannot be directly obtained from the outside of the battery using testing equipment; furthermore, the models require substantial computational resources and suffer from low efficiency. Even with simplified assumptions about the battery dimension, the complexity of P2D electrochemical models remains high, especially after coupling the battery thermal model with the battery aging reaction, significantly extending simulation time. This makes lithium-ion battery aging mechanism models unsuitable for direct application in automotive applications, but rather suitable for battery companies with access to relevant manufacturing and material parameters to simulate battery performance. Data-driven prediction models for lithium-ion battery life (SOH) often employ time-series neural network models, such as LSTM neural networks. While these models offer good predictive capabilities for time-series signals like battery life degradation, they still have shortcomings: ① Although neural network models based entirely on external battery data can accurately predict battery SOH values, their lack of identification of internal battery aging modes causes the predicted SOH degradation trajectory to deviate from the actual value in the later stages. ② For lithium-ion batteries with different historical operating conditions, their State of Harm (SOH) trajectories are drastically different. If the historical operating conditions of a lithium-ion battery are harsh, its SOH is likely to decrease rapidly in subsequent operation; if the historical operating conditions are mild, the subsequent service life of the lithium-ion battery will be longer. Data-driven models have limited ability to identify this pattern. The traditional fast charging method for lithium-ion batteries is constant current-constant voltage charging (CC-CV), which is difficult to balance charging efficiency and battery lifespan simultaneously. New fast charging methods for lithium-ion batteries have low adaptability to operating conditions, often requiring multiple calibrations at different battery operating temperatures. This results in low algorithm development efficiency and a significantly extended development cycle for charging strategies.

[0004] In summary, although the existing methods can predict the SOH value of the battery relatively accurately, the predicted SOH degradation trajectory will deviate from the actual value in the final stage because they lack the identification of the battery's internal aging mode. Summary of the Invention

[0005] To address the shortcomings of existing technologies in identifying internal aging modes of batteries, which leads to deviations in the predicted battery SOH decay trajectory from the actual value, this invention proposes a method, system, device, and storage medium for predicting the internal aging modes of lithium batteries. By measuring the battery's operating current, voltage, time, and surface temperature from external sources, the invention predicts the internal aging modes of lithium-ion batteries, thereby solving the problems existing in the prior art.

[0006] A method for predicting the internal aging mode of a lithium battery includes the following steps:

[0007] Acquire external signals from the lithium battery; the external signals include the lithium battery's voltage, current, temperature, and charging time;

[0008] A lithium battery aging mode prediction model is constructed; the lithium battery aging mode prediction model includes a lithium-ion battery aging mechanism module and a lithium-ion battery data driving module; the lithium-ion battery data driving module includes a two-layer LSTM neural network, a fully connected layer, and an attention mechanism layer;

[0009] External signals from the lithium battery are input into the lithium battery aging mode prediction model to predict the internal aging mode of the lithium battery; this specifically includes the following steps:

[0010] A mapping relationship between external signals and internal aging modes of a lithium-ion battery is established through an aging mechanism module.

[0011] By inputting the mapping relationship between the external signals of the lithium battery and its internal aging mode into the lithium-ion battery data driving module, and then passing it through the forget gate, input gate and output gate of two layers of LSTM neural network, time series data is obtained.

[0012] The attention mechanism layer integrates individual features in time-series data, and the integrated key information is output through a fully connected layer to obtain the lithium metal deposition current inside the lithium battery.

[0013] The aging mode inside a lithium battery can be predicted based on the lithium metal deposition current inside the battery.

[0014] Furthermore, the mapping relationship between external signals and internal aging modes of the battery is established through the aging mechanism module of the lithium-ion battery. Specifically, this involves establishing the mapping relationship between external signals and internal aging modes of the lithium battery through a pseudo-two-dimensional electrochemical model. This includes obtaining the electrochemical reactions that occur on the surface of the electrode active particles during lithium battery operation through the P2D reaction kinetic equation, simulating the influence of the lithium battery's own temperature changes on the internal reaction process by coupling a 0-dimensional heat transfer module into the P2D electrochemical model, and introducing the relationship between lithium battery cycle aging and lithium deposition and SEI film growth to establish the mapping relationship between external signals and internal aging modes of the lithium battery.

[0015] Furthermore, the step of inputting the mapping relationship between the external signals of the lithium battery and its internal aging mode into the lithium-ion battery data driving module and sequentially passing it through the forget gate, input gate, and output gate of two layers of LSTM neural network to obtain time-series data specifically includes the following steps:

[0016] The mapping relationship between external signals and internal aging modes of lithium batteries is used as input data and passed through a forget gate to obtain the retained data information;

[0017] The input data is controlled by the input gate. The input data and the hidden state are input into the sigmoid activation function and the tanh activation function respectively. The output of the sigmoid activation function is multiplied by the output of the tanh activation function.

[0018] The result of the multiplication is combined with the data retained by the forget gate, and then the output gate is used to predict the temporal data carried by the next hidden state.

[0019] Furthermore, the lithium battery aging mode prediction model uses the sigmoid function as the activation function, the mean squared error as the loss function, and an optimizer RMSprop that can adaptively adjust the parameter learning rate.

[0020] Furthermore, the determination of the battery aging mode based on the predicted internal lithium metal deposition current is specifically determined by whether lithium metal deposition occurs inside the battery. If no lithium metal deposition occurs, it is a single solid electrolyte passivation film growth aging mode; otherwise, it is a mixed aging mode of lithium metal deposition and solid electrolyte passivation film growth or an aging mode dominated by lithium metal deposition.

[0021] The present invention also includes a lithium battery internal aging mode prediction system, comprising:

[0022] The acquisition module is used to acquire external signals of the lithium battery; the external signals include the voltage, current, temperature, and charging time of the lithium battery.

[0023] The model building module is used to build a lithium battery aging mode prediction model; the lithium battery aging mode prediction model includes a lithium-ion battery aging mechanism module and a lithium-ion battery data driving module; the lithium-ion battery data driving module includes a two-layer LSTM neural network, a fully connected layer, and an attention mechanism layer.

[0024] The prediction module is used to input external signals from the lithium battery into the lithium battery aging mode prediction model to predict the internal aging mode of the lithium battery; it specifically includes the following steps:

[0025] The relationship establishment unit is used to establish a mapping relationship between external signals of the battery and internal aging modes of the battery through the aging mechanism module of the lithium-ion battery.

[0026] The time-series data acquisition unit is used to obtain time-series data by inputting the mapping relationship between the external signals of the lithium battery and its internal aging mode into the lithium-ion battery data driving module and passing it through the forget gate, input gate and output gate of two layers of LSTM neural network.

[0027] The integration unit is used to integrate individual features in time-series data through the attention mechanism layer, and output the integrated key information through the fully connected layer to obtain the lithium metal deposition current inside the lithium battery.

[0028] The prediction unit is used to predict the aging mode inside the lithium battery based on the lithium metal deposition current inside the lithium battery.

[0029] The present invention also includes a computer device for predicting the internal aging mode of a lithium battery, comprising: a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method for predicting the internal aging mode of a lithium battery.

[0030] The present invention also includes a readable storage medium storing a computer program, the computer program including program instructions that, when executed by a processor, are used to perform the steps of the method for predicting internal aging modes of lithium batteries.

[0031] This invention provides a method, system, device, and storage medium for predicting the internal aging mode of lithium batteries, which has the following beneficial effects:

[0032] This invention integrates a lithium-ion battery aging mechanism module with a lithium-ion battery data-driven module to construct a lithium-ion battery aging mode prediction model, enabling the prediction of specific internal aging modes of the battery. It utilizes the lithium-ion battery aging mechanism module to establish a mapping relationship between external signals during lithium-ion battery operation and internal aging modes. This mapping relationship is input into the lithium-ion battery data-driven module, allowing the LSTM neural network to learn the mapping relationship between external signals and internal aging modes. This enables the battery aging mode recognition system to accurately predict the internal aging modes of lithium-ion batteries based on external signals (charging current, charging time, battery voltage, battery temperature) during battery operation. This invention can analyze the internal aging modes of batteries based on external signal data during battery operation without disassembling the battery, and can also identify the internal aging modes of batteries during rapid charging under complex and variable environmental temperatures. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating the prediction process of the internal aging mode of a lithium battery in an embodiment of the present invention.

[0034] Figure 2 This is a schematic diagram of the mechanism-data fusion battery aging pattern recognition system in an embodiment of the present invention;

[0035] Figure 3 This is a schematic diagram of the pseudo-two-dimensional (P2D) electrochemical model in an embodiment of the present invention.

[0036] Figure 4 This is a diagram of the LSTM neural network structure in an embodiment of the present invention;

[0037] Figure 5 This is a schematic diagram of the battery life prediction data-driven model in an embodiment of the present invention;

[0038] Figure 6 This is a comparison chart of battery voltage curves from bench tests and simulation tests in an embodiment of the present invention; Figure 6 Subgraphs (a), (b), and (c) correspond to operating conditions of -10℃, 0℃, and 25℃, respectively.

[0039] Figure 7 This is a comparison chart of battery temperature curves from bench tests and simulation tests in an embodiment of the present invention; Figure 7 Subgraphs (a), (b), and (c) correspond to the temperature rise of the battery at ambient temperatures of -10℃, 0℃, and 25℃, respectively.

[0040] Figure 8 This is a schematic diagram showing the variation of lithium metal deposition current and battery capacity loss curve under different charging rates and ambient temperatures in embodiments of the present invention.

[0041] Figure 9 This is a schematic diagram showing the change of lithium metal deposition current in the battery with temperature and the capacity loss curve caused by lithium metal deposition during 3C rate charging in an embodiment of the present invention.

[0042] Figure 10 This is a schematic diagram of the simulation results given by the aging mechanism model in the embodiments of the present invention;

[0043] Figure 11 This is a schematic diagram of the battery lithium metal deposition current variation curve predicted by the mechanism-data fusion battery aging pattern recognition system in an embodiment of the present invention.

[0044] Figure 12 This is a schematic diagram of the dynamic health boundary surface of the battery in an embodiment of the present invention;

[0045] Figure 13 This is a graph showing the relationship between the battery's voltage, current, temperature, charge, and time in an embodiment of the present invention.

[0046] Figure 14 A schematic diagram of the predicted lithium metal deposition current inside the battery is provided for the mechanism-data fusion battery aging pattern recognition system in this embodiment of the invention.

[0047] Figure 15 This is a statistical chart of lithium metal deposition current conversion point data in an embodiment of the present invention;

[0048] Figure 16 This is a schematic diagram of a three-dimensional mapped surface in an embodiment of the present invention. Detailed Implementation

[0049] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0050] like Figure 1 As shown, this invention proposes a method for predicting the internal aging modes of lithium batteries. It integrates a lithium-ion battery aging mechanism module with a lithium-ion battery data-driven module to form a mechanism-data fusion battery aging mode recognition system, enabling the prediction of specific internal aging modes of the battery. A dataset of external signals and internal aging modes during lithium-ion battery operation is established using the lithium-ion battery aging mechanism module. This dataset is used to train the lithium-ion battery data-driven module, allowing the network to learn the mapping relationship between external signals and internal aging modes during lithium-ion battery operation. This enables the battery aging mode recognition system to accurately predict the internal aging modes of lithium-ion batteries based on external signals (charging current, charging time, battery voltage, battery temperature) during lithium-ion battery operation. Specifically, the method includes the following steps:

[0051] S1. Constructing a lithium-ion battery aging mechanism module; This module uses a pseudo-two-dimensional (P2D) electrochemical model for battery aging mechanisms, the basic principles of which are as follows: Figure 3 As shown.

[0052] S1.1 Constructing the P2D reaction kinetic equation: The electrochemical reactions occurring on the surface of the electrode active particles during lithium-ion battery operation can be described by the Butler-Volmer kinetic equation:

[0053]

[0054] Where, α a =α c =0.5 is the anode / cathode transfer coefficient; η is the overpotential at the solid-liquid phase contact surface, and its calculation formula is as follows:

[0055] η=φ s -φ e -E ocv -i s R film

[0056] Where, φ s With φ e These are the solid / liquid phase potentials, R. film E is the resistance generated by the solid electrolyte passivation film (SEI) on the particle surface. ocvThe open-circuit potential is determined by the lithium-ion concentration ratio on the surface of the electrode active material particles and is a quantity that depends only on the characteristics of the electrode active material itself. The exchange current density i0 is used to express the ease with which a reaction occurs at the electrode, and its specific calculation formula is as follows:

[0057]

[0058] Where k s It is the constant of electrochemical reactions; C s,MAX This represents the maximum lithium intercalation concentration of the electrode active material.

[0059] The diffusion process of lithium ions in electrode active material particles can be described by Fick's law:

[0060]

[0061] in Let represent the gradient along the radius r direction within the solid particle, and its boundary conditions are as follows:

[0062]

[0063] The above equation indicates that the lithium-ion flux is zero at the particle center; at the particle surface, the lithium-ion flux is the ion flux participating in the electrochemical reaction. In the electrolyte, lithium ions undergo both diffusion and migration processes, and their motion equations can be obtained from concentrated solution theory:

[0064]

[0065] in This represents the gradient along the electrode thickness direction x; It represents the effective diffusion coefficient of the liquid phase in a lithium-ion battery.

[0066] The change in lithium-ion concentration in the electrolyte is mainly affected by concentration diffusion and electrochemical electromigration processes. Electrolytes are present in the positive electrode, negative electrode, and separator of a lithium-ion battery. The lithium-ion concentration conditions at the corresponding boundaries are as follows:

[0067]

[0068] The above formula indicates that the lithium ion concentration is 0 and there is no lithium ion flow at the interface between the electrode and the current collector; while at the interface between the electrode and the separator, the lithium ion concentration and concentration gradient are continuous, and the lithium ion flow rates on both sides of the interface are the same.

[0069] According to Faraday's law, the ion current densities in the solid and liquid phases are respectively:

[0070] ▽i1=-a s i s =-a sFj Li

[0071] ▽i2=a s i s =a s Fj Li

[0072] in i s A positive value indicates that lithium ions are detached from the electrode active particles, i1 represents the solid phase current density, i2 represents the liquid phase current density, and i s a represents the current density of the electrochemical reaction on the surface of the active particles. s The relative specific surface area is the ratio of the surface area to the volume of a spherical particle. According to Ohm's law, the relationship between current density and electric potential in a solid phase is as follows:

[0073]

[0074] in The effective conductivity of the solid phase.

[0075] In an electrolyte, the liquid phase potential is composed of both the ion migration potential and the Ohm's law potential of the ion current:

[0076]

[0077] in The effective conductivity of the liquid phase.

[0078] In the invention, the negative current collector is selected for grounding, i.e., φ s | x＝1 =0. The liquid phase potential is continuous at the interface between the diaphragm and the electrode, and its derivative along the x-direction is also continuous. Therefore, the boundary conditions for the liquid phase potential are as follows:

[0079]

[0080] In the model, the battery is mainly divided into the anode, separator, and cathode, corresponding to the subscripts a, s, and c. Additionally, when dealing with diffusion equations, Ohm's law, and charge conservation equations, solid and electrolyte phases appear, corresponding to the subscripts s and e. To avoid confusion, unless otherwise specified, the subscript s refers to the solid phase.

[0081] S1.2 Constructing the Battery Heat Transfer Equation: To make the battery temperature more closely resemble real-world conditions during simulation and to simulate the impact of battery temperature changes on internal reaction processes, this invention couples a 0-dimensional heat transfer module into the P2D electrochemical model to simulate the battery's self-heating phenomenon. The battery's heat sources can be divided into ohmic heat, reaction heat, and polarization heat. Ohmic heat can be further divided into three parts, and its formula is as follows:

[0082]

[0083] in To solidify the ohmic heat source, As an ohmic heat source for liquid interaction, Ohmic heat generated in the direction of battery thickness due to additional resistance (current collector, etc.).

[0084] In addition to ohmic heat, the charging and discharging process of a battery also involves the heat of reaction from chemical reactions, as shown in the following formula:

[0085]

[0086] in It is the entropy coefficient of the positive and negative electrode materials, which is only related to the material itself and its lithium intercalation rate.

[0087] The polarization heat of a battery is related to the overpotential at the interface, and the formula is as follows:

[0088]

[0089] Heat dissipation occurs between the battery surface and the external environment. The formula for heat convection dissipation from the battery surface to the environment is as follows:

[0090]

[0091] Where h is the convective heat transfer coefficient, T is the battery temperature, and T amb The ambient temperature.

[0092] Considering the ohmic heat, reaction heat, polarization heat, and heat dissipation of the battery to the external environment, the temperature rise during battery charging and discharging can be expressed as:

[0093]

[0094] Where m cell c cellHere, L represents the battery's mass and specific heat capacity, respectively, and L is the thickness of the single-layer battery material. The heat of reaction, heat of polarization, and battery current in the battery thermal model are output from the battery electrochemical model. The thermal model then calculates the total heat generation and heat dissipation of the battery, providing the average battery temperature after thermal equilibrium, which is then passed to the battery electrochemical model. The battery electrochemical model needs to correct temperature-sensitive parameters in the battery electrochemical reaction after obtaining the iterated battery temperature. In the P2D model, some chemical reaction parameters (diffusion coefficient, reaction rate) change significantly with temperature; this invention uses the Arrhenius formula for correction in the model.

[0095]

[0096] Where X i The parameters that need to be corrected, For its reference temperature T ref The value of E i The corresponding parameter is the activation energy of the reaction, and T is the battery temperature. The correction factors required in this paper are: solid-phase diffusion coefficient of positive and negative electrodes, liquid-phase diffusion coefficient, liquid-phase conductivity, and chemical reaction rate coefficient of positive and negative electrodes.

[0097] S1.3 Constructing the Battery Aging Equation: To understand the relationship between lithium battery cycle aging and lithium deposition and SEI film growth in this invention, correction terms for these two reactions are introduced into the lithium-ion battery electrochemical model. These are side reactions that compete with the normal lithium-ion intercalation reaction during battery charging and discharging, consuming recyclable lithium ions in the electrolyte and causing a decrease in usable battery capacity. In this battery aging model, two main side reactions are considered: lithium metal deposition and solid electrolyte membrane growth. These two side reactions compete with the lithium-ion intercalation reaction during battery operation, thus the complete current density can be expressed as:

[0098] i tot =i Li +i loc,pla +i loc,SEI

[0099] Where i tot i is the total current density through the solid-liquid interface. Li Lithium-ion flux density at the surface of active particles in electrochemical reaction, i loc,pla Local current density of lithium metal deposition side reaction, i loc,SEI Local current density of side reactions during solid electrolyte membrane growth.

[0100] Lithium metal deposition occurs because lithium ions at the solid-liquid interface cannot be promptly embedded into the graphite particles of the negative electrode. This accumulation of lithium ions causes a change in local potential, reaching the potential required for lithium ion reduction to metallic lithium. This phenomenon typically occurs in low-temperature operating environments or under high-rate charging conditions. However, as the battery nears the end of its lifespan, the electrode surface becomes fragmented and rough, making it easier for local overpotentials to occur, leading to lithium metal deposition. Lithium metal deposition in lithium-ion batteries can be described by the Butler-Volmer kinetic equations:

[0101]

[0102] η pla =φ s -φ e -E eq,pla -i s R film

[0103] Where k pla η represents the lithium metal deposition reaction rate. pla E is the overpotential for lithium metal deposition. eq,pla C is the equilibrium potential for the lithium metal deposition reaction. s,surf This represents the lithium-ion concentration on the surface of the solid particles. After lithium metal is deposited, some of the lithium ions that are converted into metallic lithium no longer participate in the lithium-ion cycle, resulting in a reduction in the cyclic lithium in the battery.

[0104] Local current density i of SEI film growth side reactions loc,SEI It can be represented as:

[0105]

[0106] η SEI =φ s -φ e -E eq,SEI -i s R film

[0107] Where i loc,1C,ref η is the local current density during 1C discharge. HK is the dimensionless graphite expansion factor, a function of the graphite's state of charge, which is 0 during lithium-ion intercalation / deintercalation. SEI q represents the overpotential during SEI film growth. SEI To facilitate charge accumulation during the formation of the solid electrolyte SEI film, E eq,SEI This is the equilibrium voltage for the SEI film growth reaction.

[0108] Unlike metallic lithium, which conducts electrons, the SEI film cannot conduct electrons and only allows lithium ions to pass through. The presence of the SEI film hinders lithium ion transport, and its impact on film resistance during growth needs to be calculated. The formulas for the change in SEI film thickness and internal resistance are as follows:

[0109]

[0110]

[0111] Where δ is the SEI film thickness, M SEI ρ represents the relative mass of the SEI membrane. SEI κ represents the density of the SEI film. SEI The conductivity of the SEI film is given. The growth of the SEI film not only consumes active lithium ions but also increases the internal membrane resistance of the battery, affecting the battery's energy density and power density.

[0112] S2. Construct a lithium-ion battery data-driven module.

[0113] S2.1 Constructing LSTM Neural Network Units; Since the data from lithium-ion batteries is collected during charging and discharging cycles, it belongs to time-series data. Recurrent Neural Networks (RNNs) are widely used to process time-series data. However, RNNs suffer from gradient explosion or vanishing problems, which make the model difficult to train and optimize. To solve this problem, the Long Short-Term Memory (LSTM) network structure was proposed, which uses an LSTM recurrent unit structure to replace the state units of the classic RNN. Each unit of LSTM is as follows: Figure 4 As shown.

[0114] Building upon the traditional RNN's memory units used to store all information, the LSTM network introduces a gate mechanism to control the flow and loss of features. At each time step, information input from the input layer first passes through the input gate; the opening and closing of the input gate determines whether information is input into the memory unit at that moment. Then, at each time step, the values ​​in the memory unit undergo a forgetting process, controlled by the forget gate. The LSTM unit... Figure 4 As shown in (a) in the figure.

[0115] Training data using an LSTM unit can be divided into four steps. The first step involves passing through a forget gate, which controls the discarding and retention of information, such as... Figure 4 As shown in (b) above. When the LSTM unit obtains x t and h t-1 When processing data, the sigmoid activation function outputs a vector containing either 0 or 1, which determines the cell state C. t-1Which information is discarded or retained? In the vector, 0 represents forgetting and 1 represents retention. The forget gate can be formulated as:

[0116] f t =σ(W f x t +W f h t-1 +b f )

[0117] In the second step, the input information is controlled by the input gate. To update the LSTM cell state, an input gate is required, such as... Figure 4 As shown in (c), the input gate consists of two parts: a sigmoid activation function and a tanh activation function. These two parts participate in C as a factor and a new state, respectively. t Update. First, this invention passes the previous hidden state and the current input to the sigmoid activation function, whose output vector of 0s or 1s determines which values ​​to update; 0 indicates insignificant, and 1 indicates significant. Simultaneously, the hidden state and the current input are passed to the tanh activation function, compressing them to between -1 and 1 to aid in network tuning. Then, the outputs of the sigmoid and tanh activation functions are multiplied. Their calculations are as follows:

[0118] i t =σ(W i x t +W i h t-1 +b i )

[0119]

[0120] The third step is to combine the data from the forget gate and the input gate. Then, the cell state changes from C. t-1 Updated to C t :

[0121]

[0122] Finally, the fourth step is to output the gate, such as... Figure 4 As shown in (d) in the diagram. The output gate determines the next hidden state. The hidden state contains information about the previous input and is also used for prediction. First, the previous hidden state and the current input are passed to the sigmoid activation function, and then the new cell state is passed to the tanh activation function. Finally, the output of the sigmoid activation function is multiplied by the output of the tanh activation function to determine the hidden state h. t The information to be carried is expressed in the following formula:

[0123] o t =σ(Wo x t +W o h t-1 +b o )

[0124] h t =o t *tanh(C t )

[0125] In the four steps described above, W represents the weight of the corresponding gate, and b represents the bias term of the corresponding gate. By introducing input gates, forget gates, and output gates, LSTM can effectively control the flow of information and retain important historical information, thereby improving the performance and stability of the model.

[0126] S2.2 Setting up an attention mechanism; The attention mechanism is a machine learning technique that helps neural networks focus more on important features, thereby improving model performance and generalization ability. In neural networks, each feature has a different impact on the result, but usually only one set of features dominates the output. The attention mechanism learns based on the attention level of individual features in the sequence and integrates features according to that attention level. Even if there is only one feature input to the LSTM network, when processing time-series data, the attention mechanism can still capture key information in the time-series data better by weighting different time steps. The attention mechanism is expressed as follows:

[0127] u t =tanh(W u *h t +b u )

[0128]

[0129] s t =a t *h t

[0130] Where W is the weight; b represents the bias term; a represents the weight of each attribute; and s represents the prediction result after weighted summation.

[0131] Because the difference operations during IC feature data processing amplify measurement noise, and filtering is insufficient to eliminate the impact of noise on training results, abnormal capacity spikes occur when the model predicts SOH. Therefore, an attention mechanism layer is introduced into the LSTM neural network. This allows the LSTM model to automatically adjust the value and position of the IC curve peak and the weight allocation of the subsequently added auxiliary feature KPP during learning. Simultaneously, the dependencies between each feature and time step are optimized, thereby improving the model's ability to resist abnormal capacity spikes and its transfer and generalization capabilities.

[0132] S2.3, Building the Neural Network; The battery life prediction data-driven model built in this invention uses a two-layer LSTM neural network, such as... Figure 5 As shown, each LSTM neural network layer consists of 32 LSTM units. After data passes through each layer, a forgetting operation is performed to prevent overfitting; the forgetting rate for each LSTM neural network layer is 0.5. Finally, an attention mechanism layer is added between the second LSTM neural network layer and the fully connected layer to help the neural network adjust feature weights, thereby improving the prediction performance and generalization ability of the battery life prediction model. This invention chooses the basic sigmoid function as the model's activation function, used as a gating switch for data transmission. Like RNNs, LSTM uses gradient descent to learn from the data, therefore a loss function is needed as the target for learning or network optimization.

[0133] In this invention, Mean Squared Error (MSE) is chosen as the loss function. To improve learning efficiency, an optimizer, RMSprop, which adaptively adjusts the learning rate of each parameter, is employed. The RMSprop optimizer maintains an exponentially weighted average in each iteration to adjust the learning rate of each parameter in the network. If the gradient of a parameter is large, the RMSprop algorithm automatically decreases its learning rate; if the gradient is small, it increases the learning rate, thereby improving training efficiency and enabling the model to converge faster.

[0134] S3, Aging Mechanism Module Verification and Simulation Dataset Generation.

[0135] S3.1 Verification of the aging mechanism module: When building the mechanism-data fusion battery aging pattern recognition system, the aging mechanism module must first be calibrated so that its output results are close to the data collected by the bench test. In this way, the data obtained by the mechanism module simulation can be used to replace the bench test data to train the data-driven module, thereby reducing the time cost and workload of the data-driven module training data collection.

[0136] During the invention process, the inventors selected battery voltage and temperature data to verify the accuracy of the output results of the corrected mechanism module. Battery voltage data represents the external electrical signal during battery operation and is also an important battery performance characteristic; while battery temperature has a significant impact on the internal mass transport process and is one of the decisive factors in the battery's aging mode. Different temperature conditions were also selected to further verify the model's generalization ability under multiple temperature conditions. The verification process included battery charge-discharge bench experiments at -10℃, 0℃, and 25℃, as well as simulation experiments under the corresponding conditions.

[0137] like Figure 6The figure shown is a comparison of the battery voltage curves from the bench test and the simulation test. Figure 6 (a), (b) and (c) correspond to operating conditions of -10℃, 0℃ and 25℃, respectively. Figure 6 The darker curve represents the voltage curve output from the mechanism module simulation, while the lighter curve represents the voltage curve acquired from the bench test. It can be observed that at low temperatures, the constant current charging range of the battery is shorter, while the constant voltage charging range is longer, with a high degree of overlap between the two voltage curves in the constant current range. At room temperature (25℃), due to the improved diffusion and transport capabilities of materials within the battery, lithium ions can be inserted into the negative electrode particles more quickly, so the time for the battery terminal voltage to reach 4.2V is much longer than under low-temperature conditions.

[0138] like Figure 7 The figure shown is a comparison of the battery temperature curves from bench tests and simulation tests. Figure 7 Subgraphs (a), (b), and (c) correspond to the temperature rise of the battery at ambient temperatures of -10℃, 0℃, and 25℃, respectively. Figure 7 The darker curve represents the temperature curve output from the mechanism module simulation, while the lighter curve represents the temperature curve collected from bench experiments. During charging, the battery temperature initially rises and then slowly decreases. When the battery is in the constant current charging stage, the high current causes a rapid rise in battery temperature, especially at -10℃. Due to the high internal resistance of the battery, a significant temperature rise occurs, bringing its maximum temperature close to that of a battery operating at 0℃. This temperature rise greatly improves battery performance, which explains the similar charging times at -10℃ and 0℃. When the battery enters the constant voltage charging stage, the charging current gradually decreases, the battery's self-heating value is less than its heat dissipation value, and the battery temperature begins to decrease.

[0139] Overall, the simulation results provided by the mechanism module are close to the actual data from bench experiments. The simulation results from the mechanism module can be used to train the data-driven module instead of the bench experiment results, improving the efficiency of model training data collection. Furthermore, the mechanism module can provide overpotential at the negative electrode-separator and lithium metal deposition current signals that cannot be directly collected inside the battery, enabling the data-driven module to identify the aging mode inside the battery based on the acquired external signal characteristics.

[0140] S3.2 Simulation Dataset Generation: To enable the mechanism-data fusion battery aging pattern recognition system model to identify battery aging patterns, aging-related parameter signals need to be added. Previous research has shown that the growth of the solid electrolyte film inside the battery does not fluctuate drastically with operating conditions, only changing slightly with the battery's operating temperature; however, lithium metal deposition rapidly consumes recyclable lithium ions in the electrolyte, causing a sharp drop in usable battery capacity. Furthermore, the growth of the solid electrolyte passivation film is unavoidable during battery use; its growth rate can only be slowed down as much as possible. Conversely, lithium metal deposition can be prevented by adjusting the battery's operating state. Therefore, when using a BMS for refined management of the entire battery lifecycle, greater attention is paid to whether aging reactions such as lithium metal deposition occur inside the battery, which can lead to rapid performance degradation. When establishing the battery simulation dataset, the focus is also on the lithium metal deposition reaction inside the battery, obtaining the reaction current of lithium metal deposition inside the battery through the mechanism module. If the simulated current for lithium metal deposition is 0A, then no lithium metal deposition occurs inside the battery; if it is not 0A, then the amount of charge obtained by integrating the simulated current for lithium metal deposition over time is positively correlated with the amount of lithium ions consumed by the lithium metal deposition side reaction.

[0141] In battery charging simulations, the overpotential at the negative electrode-separator is used to characterize whether the lithium metal deposition reaction has reached its threshold, while the lithium metal deposition reaction current expresses whether the lithium deposition reaction occurs and its intensity. The lithium metal deposition reaction current is highly correlated with the simulation conditions; temperature and charging current rate both have a significant impact. Since the battery charging time varies under different charging rates and ambient temperatures, the state of charge (SOC) is used as the horizontal axis of the graph to more intuitively compare the impact of the simulation conditions on the lithium metal deposition side reactions. Figure 8 The figure shows the variation of lithium metal deposition current at different charging rates at 0°C, and the capacity loss caused by lithium metal deposition. The red curve represents 5C charging, the green curve represents 3C charging, and the red curve represents 1C charging. Figure 8 As can be seen in (a), the lithium metal deposition current initially increases rapidly, then decreases rapidly when the charging mode changes from constant current to constant voltage. Furthermore, the lithium metal deposition reaction current at a 5C charging rate is significantly higher than that of batteries charged at lower rates. Correspondingly... Figure 8 (b) shows the battery capacity loss caused by lithium metal deposition, and the results are consistent with the changes in lithium deposition current.

[0142] like Figure 9 The figure shows the variation of lithium metal deposition current within the battery with temperature during 3C charging, as well as the capacity loss caused by lithium metal deposition. In the figure, the blue curve represents -10℃, the green curve represents 0℃, the red curve represents 10℃, and the cyan curve represents 25℃. Figure 9As can be seen from (a), in terms of lithium metal deposition side reactions, the effects of decreasing ambient temperature and increasing charging current rate on battery operation are similar. Figure 9 The capacity loss caused by lithium deposition in (b) is also related to Figure 9 The side reaction current corresponds to (a) in the equation.

[0143] Currently, the maximum charging current rate of electric vehicles on the market is mainly concentrated at 4C or 5C. Considering non-fast charging usage scenarios, this invention sets the simulated battery charging rate range to 1C to 5C, corresponding to the battery capacity of the simulated object, i.e., a charging current of 5A to 25A. Considering the usage scenarios of cold starts in northern winters, or charging after long periods of parking at low temperatures, as well as battery aging under multiple temperature conditions, this invention sets the simulation temperature range to -20℃ to 20℃, with a set of simulation experiments set at 10℃ intervals. A total of 25 simulation conditions are conducted, as shown in Table 1. These data will be used as the test set and validation set in the battery aging mode prediction, respectively, to determine the estimation accuracy of the mechanism-data collaboratively driven battery aging model for the internal lithium metal deposition reaction of the battery.

[0144] Table 1 Simulation Experiment Data

[0145]

[0146] like Figure 10 The figure shows the simulation results given by some aging mechanism models. Figure 10 The curve in the middle represents the lithium metal deposition current inside the battery, with the vertical axis representing the current magnitude and the horizontal axis representing time. Observing the curve can reveal the time and intensity of the lithium metal deposition reaction.

[0147] S4, Mechanism-Data Fusion Battery Aging Pattern Recognition System Training and Prediction, such as Figure 2 As shown.

[0148] S4.1 Training of the Aging Pattern Recognition System: During system training, all operating conditions at 0℃ and 3C charging were selected as the training set, and the remaining data were used as the test set. The ratio of the training set to the validation set was 9:16, and both the training set and the test set included operating conditions where lithium metal deposition occurred and those where it did not, which can fully train / detect the model's ability to predict side reactions. The selection of the training set is shown in Table 2.

[0149] Table 1 Training Data Selection Table

[0150]

[0151] Note: "Training" indicates the neural network training set, and "Test" indicates the neural network test set; unbolded indicates that lithium metal deposition occurred during the charging process, while bold indicates that no lithium metal deposition occurred throughout the entire process.

[0152] S4.1, Mechanism – Battery aging pattern recognition system predicts lithium metal deposition current as follows: Figure 11 As shown. Figure 11 The light-colored curve represents the actual value of lithium metal deposition current in the test set, while the dark-colored curve represents the predicted value of lithium metal deposition current given by the model. No lithium metal deposition occurred inside the battery charged at a 1C current rate, and the system's prediction result is also a flat curve. Similarly, no lithium metal deposition occurred in batteries charged at any rate at 20℃, and the prediction result is also a straight line. For batteries without lithium metal deposition, the mechanism-data fusion battery aging pattern recognition system predicts the internal lithium metal deposition current with 100% accuracy, accurately predicting the battery aging model under charging conditions. Batteries charged at 1C current rate at 10℃ and batteries charged at 1C–5C current rates at 20℃ both exhibit a single solid electrolyte film growth aging mode.

[0153] Under test conditions where lithium metal deposition occurs, the degree to which the model-predicted lithium metal deposition current approximates the actual value is somewhat correlated with the magnitude of the lithium metal deposition current itself. (Observation) Figure 11 The six subplots (d), (j), (h), (j), (k), and (i) in the model show a high degree of overlap between the predicted curves and the actual values, with similar curve shapes. Furthermore, they also demonstrate good predictive performance for the initiation and termination points of the lithium metal deposition reaction.

[0154] To better illustrate the prediction performance of each test set, Table 3 presents the prediction error RMSE and MAE results, along with the normalized RMSE, facilitating horizontal comparisons under different lithium metal deposition current conditions. Since lithium metal deposition does not occur at any charging rate at 20°C and at a 1C charging current rate at 10°C, the predicted lithium metal deposition current given by the mechanism-data-driven co-estimation model is consistently 0A. Therefore, the errors for the above five conditions are all 0A. For batteries with a constant lithium metal deposition current of 0A, the capacity decay under the corresponding conditions is due to the slow growth of the solid electrolyte membrane. Under this aging mode, lithium-ion batteries have a longer lifespan.

[0155] Table 3 Prediction Error RMSE and MAE Results

[0156]

[0157]

[0158] For charging current rates where lithium metal deposition occurs at an ambient temperature of 10°C, the true value of the lithium metal deposition current in the test data is relatively small, resulting in relatively low RMSE and MAE estimation errors. To better compare the prediction results of the mechanism-data-driven co-estimation model under different charging current rates, this invention normalizes the RMSE of the prediction results to eliminate the influence of current dimensions. The normalized RMSE results show that the estimation errors of the internal lithium metal deposition current for batteries charged at 2C and 3C at 10°C are very large, at 47% and 59% respectively, far exceeding the 19% error under the 5C charging rate condition.

[0159] S5, for extracting dynamic health boundaries for fast charging.

[0160] In the simulation of lithium metal deposition current in a battery, this invention can obtain the functional relationship between the lithium metal deposition current inside the battery and the battery temperature, battery SOC, and battery charging current under 25 initial operating conditions. For each operating condition where lithium metal deposition occurs, there will be at least one battery state point where the lithium metal deposition current changes from 0 to non-zero (or from non-zero to 0). The battery temperature, battery SOC, and the charging current at that point are extracted (charged in a constant current-constant voltage manner, the battery charging current will vary in the constant voltage range). After interpolation, the data is plotted as the battery dynamic health boundary surface, as shown below. Figure 12 As shown.

[0161] Embodiments of the present invention:

[0162] 1. Battery Operating Signal Acquisition: During battery charging, the charging station acquires battery voltage, current, temperature, and charging time through the Battery Management System (BMS). In the initial charging stage, the system reads the battery voltage and, through the potential mapping relationship between battery voltage and state of charge (SOC), obtains the initial battery capacity. Then, the battery capacity is calculated by integrating the charging current and time. Finally, the battery voltage, current, temperature, and capacity are plotted as a time curve for input into a mechanism-data fusion battery aging pattern recognition system, such as... Figure 13 As shown.

[0163] 2. Mechanism-Data Fusion Battery Aging Pattern Recognition System Predicts Battery Aging Patterns: Based on the input battery operating signals, the mechanism-data fusion battery aging pattern recognition system provides a predicted result for the internal lithium metal deposition current of the battery, such as... Figure 14 As shown.

[0164] Based on the predicted lithium metal deposition current, it can be determined whether lithium metal deposition has occurred inside the battery, and then it can be determined whether the battery is in a single solid electrolyte passivation film growth aging mode, a mixed aging mode of lithium metal deposition and solid electrolyte passivation film growth, or an aging mode dominated by lithium metal deposition.

[0165] 3. Plotting the dynamic health boundary for fast battery charging: Based on the predicted operating conditions where lithium metal deposition occurs, there will be at least one battery state point where the lithium metal deposition current changes from 0 to non-zero (or from non-zero to 0), referred to as the lithium metal deposition current transition point. The battery temperature, battery SOC, and the charging current at that point are extracted (using a constant current-constant voltage charging method, the battery charging current will vary within the constant voltage range), such as... Figure 15 As shown, this is used to draw the dynamic health boundary surface of the battery.

[0166] according to Figure 15 Data points in the data can be used to plot 3D mapped surfaces, such as... Figure 16 As shown, the X, Y, and Z axes represent the battery charging current, battery state of charge (SOC), and battery temperature, respectively. The outer region of the curved surface represents the operating condition where lithium metal deposition does not occur, and the battery aging mode is dominated by lithium metal deposition, or a mixed aging mode of lithium metal deposition and solid electrolyte passivation film growth. The region enclosed by the curved surface and the coordinate plane represents the region where lithium metal deposition does not occur, and the battery aging mode is solely solid electrolyte passivation film growth. The curved surface represents the maximum charging current at the corresponding battery SOC and battery temperature, specifically for the growth of a single solid electrolyte passivation film. This maximum charging current is also one of the control targets for charging settings during fast charging.

[0167] In the application of this curved surface in battery fast charging management, the maximum charging current at the given battery temperature and charge level can be calculated by interpolation using data obtained from the charging infrastructure. This current is then used as the control target for charging settings, increasing the fast charging current rate and shortening charging time without causing significant battery capacity aging, thus achieving efficient and healthy battery charging management.

[0168] This invention analyzes the internal aging side reactions of a battery, specifically lithium metal deposition, using a battery aging mechanism module within a mechanism-data fusion battery aging pattern recognition system. This determines the battery's aging mode under corresponding operating conditions, allowing analysis of the internal aging mode based on external signal data during battery operation without disassembling the battery. By fusing a battery aging mechanism model (P2D model) and a data-driven model (LSTM network), the internal aging mode of a lithium-ion battery during operation is estimated based on collected external operating signals. This enables the identification of internal aging modes during rapid charging under complex and variable environmental temperatures. Utilizing the battery aging mode judgment results provided by the mechanism-data fusion battery aging pattern recognition system, a dynamic health boundary for fast charging is extracted, balancing battery charging efficiency and lifespan. This allows for rapid charging without a significant decrease in battery lifespan.

[0169] Based on the same inventive concept, this invention proposes a lithium battery internal aging mode prediction system, comprising:

[0170] The acquisition module is used to acquire external signals of the lithium battery; the external signals include the voltage, current, temperature, and charging time of the lithium battery.

[0171] The model building module is used to build a lithium battery aging mode prediction model. The lithium battery aging mode prediction model includes a lithium-ion battery aging mechanism module and a lithium-ion battery data-driven module. The lithium-ion battery data-driven module includes a two-layer LSTM neural network, a fully connected layer, and an attention mechanism layer.

[0172] The prediction module is used to input external signals from the lithium battery into the lithium battery aging mode prediction model to predict the internal aging mode of the lithium battery; the prediction module specifically includes:

[0173] The relationship establishment unit is used to establish a mapping relationship between external signals of the battery and internal aging modes of the battery through the aging mechanism module of the lithium-ion battery.

[0174] The time-series data acquisition unit is used to obtain time-series data by inputting the mapping relationship between the external signal of the lithium battery and its internal aging mode into the lithium-ion battery data driving module and passing it through the forget gate, input gate and output gate of two layers of LSTM neural network.

[0175] The integration unit is used to integrate individual features in time-series data through the attention mechanism layer, and output the integrated key information through the fully connected layer to obtain the lithium metal deposition current inside the lithium battery.

[0176] The prediction unit is used to predict the aging mode inside the lithium battery based on the lithium metal deposition current inside the lithium battery.

[0177] Based on the same inventive concept, this invention also proposes a computer device for predicting the internal aging mode of a lithium battery, comprising: a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method for predicting the internal aging mode of a lithium battery.

[0178] Based on the same inventive concept, the present invention also proposes a readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, are used to perform steps for a method for predicting the internal aging mode of a lithium battery.

[0179] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting the internal aging mode of a lithium battery, characterized in that, Includes the following steps: Acquire external signals from the lithium battery; the external signals include the lithium battery's voltage, current, temperature, and charging time; A lithium battery aging mode prediction model is constructed; the lithium battery aging mode prediction model includes a lithium-ion battery aging mechanism module and a lithium-ion battery data driving module; the lithium-ion battery data driving module includes a two-layer LSTM neural network, a fully connected layer, and an attention mechanism layer; External signals from the lithium battery are input into the lithium battery aging mode prediction model to predict the internal aging mode of the lithium battery; this specifically includes the following steps: The aging mechanism module of lithium-ion battery establishes a mapping relationship between external signals and internal aging modes of the battery; the data obtained by simulation of the aging mechanism module of lithium-ion battery is used to train the lithium-ion battery data driving module. Time-series data is obtained by inputting external signals from the lithium battery into the trained lithium-ion battery data driving module and passing them sequentially through the forget gate, input gate, and output gate of two layers of LSTM neural network. The attention mechanism layer integrates individual features in time-series data, and the integrated key information is output through a fully connected layer to obtain the lithium metal deposition current inside the lithium battery. The aging mode inside a lithium battery can be predicted based on the lithium metal deposition current inside the battery.

2. The method for predicting the internal aging mode of a lithium battery according to claim 1, characterized in that, The process of establishing a mapping relationship between external signals and internal aging modes of lithium-ion batteries through the aging mechanism module specifically involves establishing this mapping relationship through a pseudo-two-dimensional electrochemical model. This includes obtaining the electrochemical reactions occurring on the surface of electrode active particles during lithium battery operation using P2D reaction kinetic equations, simulating the influence of lithium battery temperature changes on internal reaction processes by coupling a 0-dimensional heat transfer module into the P2D electrochemical model, and introducing the relationship between lithium battery cycle aging and lithium deposition and SEI film growth to establish the mapping relationship between external signals and internal aging modes of lithium batteries.

3. The method for predicting the internal aging mode of a lithium battery according to claim 1, characterized in that, The process of obtaining time-series data by inputting external signals from the lithium battery into the trained lithium-ion battery data driving module and passing them sequentially through the forget gate, input gate, and output gate of a two-layer LSTM neural network includes the following steps: The mapping relationship between external signals and internal aging modes of lithium batteries is used as input data and passed through a forget gate to obtain the retained data information; The input data is controlled by the input gate. The input data and the hidden state are input into the sigmoid activation function and the tanh activation function respectively. The output of the sigmoid activation function is multiplied by the output of the tanh activation function. The result of the multiplication is combined with the data retained by the forget gate, and then the output gate is used to predict the temporal data carried by the next hidden state.

4. The method for predicting the internal aging mode of a lithium battery according to claim 1, characterized in that, The lithium battery aging mode prediction model uses the sigmoid function as the activation function, the mean squared error as the loss function, and an optimizer RMSprop that can adaptively adjust the parameter learning rate.

5. The method for predicting the internal aging mode of a lithium battery according to claim 1, characterized in that, The battery aging mode is determined based on the predicted lithium metal deposition current inside the battery. Specifically, the aging mode is determined by whether lithium metal deposition occurs inside the battery. If no lithium metal deposition occurs, it is a single solid electrolyte passivation film growth aging mode; otherwise, it is a mixed aging mode of lithium metal deposition and solid electrolyte passivation film growth or an aging mode dominated by lithium metal deposition.

6. A lithium battery internal aging mode prediction system, characterized in that, include: The acquisition module is used to acquire external signals of the lithium battery; the external signals include the voltage, current, temperature, and charging time of the lithium battery. The model building module is used to build a lithium battery aging mode prediction model; the lithium battery aging mode prediction model includes a lithium-ion battery aging mechanism module and a lithium-ion battery data driving module; the lithium-ion battery data driving module includes a two-layer LSTM neural network, a fully connected layer, and an attention mechanism layer. The prediction module is used to input external signals from the lithium battery into the lithium battery aging mode prediction model to predict the internal aging mode of the lithium battery; the prediction module specifically includes: The relationship establishment unit is used to establish a mapping relationship between external signals and internal aging modes of the battery through the aging mechanism module of the lithium-ion battery; the data obtained by simulation through the aging mechanism module of the lithium-ion battery is used to train the lithium-ion battery data driving module. The time-series data acquisition unit is used to obtain time-series data by inputting external signals from the lithium battery into the trained lithium-ion battery data driving module and passing them sequentially through the forget gate, input gate, and output gate of two layers of LSTM neural network. The integration unit is used to integrate individual features in time-series data through the attention mechanism layer, and output the integrated key information through the fully connected layer to obtain the lithium metal deposition current inside the lithium battery. The prediction unit is used to predict the aging mode inside the lithium battery based on the lithium metal deposition current inside the lithium battery.

7. A computer device for predicting the internal aging mode of a lithium battery, characterized in that, include: A memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the lithium battery internal aging mode prediction method according to any one of claims 1-5.

8. A readable storage medium, characterized in that, The readable storage medium stores a computer program, the computer program including program instructions, which, when executed by a processor, are used to perform the steps of the method for predicting the internal aging mode of a lithium battery as described in any one of claims 1-5.

Images