A charging optimization method for inhibiting lithium plating based on a mechanism model

Abstract

The application claims a charging optimization method for inhibiting lithium plating based on a mechanism model, belonging to the technical field of battery charging. The method first constructs a lithium ion battery electrochemical model P2D model, and linearizes it as a state space of the model predictive control method to obtain the response of the lithium plating overpotential to the charging current, which is one of the main factors promoting battery aging. Then, a multi-objective optimization task of fast charging and limiting the size of lithium plating potential is designed, which can maximize the charging current while inhibiting the lithium plating reaction in the lithium battery, thereby realizing the fast charging scheme of inhibiting the lithium plating reaction in the battery. In addition, by constructing an LSTM model, the current and voltage data of the battery are used to identify the battery aging condition, and the experimental results show that the charging strategy can improve the charging speed while keeping the lithium plating potential of the battery within a certain range, and can effectively adjust the battery with different degrees of aging, significantly improving the universality of the charging strategy.

Description

Technical Field

[0001] This invention belongs to the field of charging strategies and is a charging optimization method based on a mechanism model to suppress lithium plating. Background Technology

[0002] In recent years, fast charging technology has attracted much attention from electric vehicle users due to concerns about insufficient driving range and long charging times. While increasing the charging current can shorten charging time, it may lead to irreversible capacity degradation or even battery failure. To balance charging efficiency and battery durability, this study designs a charging scheme that considers both charging speed and lifespan protection based on the battery aging mechanism. This scheme dynamically adjusts the charging current by constructing a state-space model of the lithium plating overpotential and combining multi-objective optimization and model predictive control techniques. Furthermore, addressing the limitations of traditional charging strategies in adapting to battery aging states, this study extracts core aging-related parameters from an electrochemical model and uses an LSTM network to train aging data, achieving real-time identification of battery state and model updates. This method, which integrates mechanistic models and data-driven approaches, not only slows down battery performance degradation while increasing charging speed but also dynamically optimizes the charging strategy based on the degree of battery aging, achieving a balance between battery lifespan and charging efficiency.

[0003] A search revealed application publication number CN116683069A, which discloses a dynamic optimization charging method for improving the active lifespan of lithium-ion batteries in the field of battery charging technology. The method includes the following steps: constructing a battery electrochemistry-heat generation-aging coupled model, and establishing quantitative mathematical relationships within this model based on a battery heat generation sub-model, a voltage sub-model, and a battery aging sub-model; based on the proposed model, and combined with a Kalman filter algorithm, constructing an accurate comprehensive state observer; accurately estimating unmeasurable internal state quantities within the battery based on measurable state quantities to accurately track the current state of the battery, and performing closed-loop state correction on some state parameters within the model; and designing an optimized charging controller based on the proposed model and state observer, combined with a model predictive control algorithm.

[0004] The method described in patent application CN116683069A modifies the state space using a Kalman filter, requiring real-time calculations of data throughout the battery's lifecycle to ensure model accuracy. This results in a heavy computational burden, especially when frequent updates are needed, and may be unsuitable for resource-constrained devices. Furthermore, closely monitoring voltage and current throughout the battery's lifecycle not only increases the complexity of data processing but may also impact processing speed due to the massive amount of data.

[0005] In contrast, this invention uses an LSTM model to periodically update the state space, focusing only on voltage and current data at the end of the charging process for parameter identification. This avoids real-time capture of data throughout the entire lifecycle, significantly reducing the computational burden. By performing parameter identification over a longer period, this method lowers the demand for computing resources, enabling the device to operate efficiently with lower power consumption and computing power. Furthermore, this method can be processed in the cloud, promoting its widespread adoption in practical applications. Summary of the Invention

[0006] This invention aims to solve the problems of the prior art mentioned above. It proposes a charging optimization method based on a mechanistic model to suppress lithium plating. The technical solution of this invention is as follows:

[0007] A charging optimization method for suppressing lithium plating based on a mechanism model includes the following steps:

[0008] Step 1: Simplify and linearize the battery electrochemical model;

[0009] Step 2: Based on the simplified battery electrochemical model, design the state space of the model predictive control method and obtain the response of the battery lithium plating overpotential to the charging current.

[0010] Step 3: By limiting the overpotential of lithium-ion battery plating and accelerating the charging speed, a multi-objective optimization task is designed, and the quadratic programming method is used to solve the battery charging current under the constraints.

[0011] Step 4: Using the LSTM model, identify battery aging parameters through battery current and voltage data, and update the state-space equations in the model predictive control accordingly, so that the charging strategy can dynamically adapt to the battery aging state.

[0012] Furthermore, Step 1, which simplifies and linearizes the battery electrochemical model, specifically includes assuming that the concentration distribution inside the particles is a parabolic function and combining it with the solid-phase concentration description of the single-particle model, which leads to the following:

[0013]

[0014]

[0015] In the formula, J_n is the orifice wall flux, and C avg C represents the average lithium-ion concentration in solid-phase particles. surf Represents the surface lithium ions of solid particles, where R represents the radius of the solid particle, and D... eff denoted by , where J represents the solid-phase diffusion coefficient and J represents the concentration flux of the lithium-ion intercalation reaction.

[0016] Furthermore, in Step 2, a model predictive control method is constructed based on the battery electrochemical model, and its state-space matrix related to the lithium-ion battery overpotential is as follows:

[0017] x(k+1)=Ax(k)+Bi(k)+C

[0018] y(k+1)=Dx(k+1)

[0019] Where x(k) is the value of the state space at time k, x = [C avg C surf JU ocp U sei η] T C avg C represents the average lithium-ion concentration in solid-phase particles. surf U represents the surface lithium-ion concentration of the solid particles, J represents the concentration flux of the lithium-ion intercalation reaction, and U represents the concentration flux of the lithium-ion intercalation reaction. ocp U represents the battery open-circuit voltage. sei Let represent the voltage drop caused by the SEI layer of the battery, η represent the overpotential of the lithium-ion battery intercalation reaction, i(k) represent the charging current at time k, i.e., the control target, and y(k) represent the lithium plating overpotential at time k. The matrix A is:

[0020]

[0021] Where R represents the radius of the solid-phase particle, and D eff C represents the solid-phase diffusion coefficient. max This represents the maximum solid-phase lithium intercalation concentration, where 'a' is a first-order term in the fitting function relating the open-circuit potential of the electrode material to the amount of lithium intercalation, and its matrix B is:

[0022]

[0023] Where A cell a represents the area of ​​the battery electrodes. s R represents the specific surface area of ​​the battery material, L represents the electrode length, F represents the Faraday constant, and R represents the electrode length. sei R represents the SEI layer film resistance of the battery. ct Represents charge transfer impedance, and its C matrix is:

[0024] C = [0 b 0 0 0 0] T

[0025] Where b is a constant term in the function relating the open-circuit potential of the electrode material to the amount of lithium intercalation, the D matrix is ​​represented as follows:

[0026] D = [0 0 0 1 1 1] T .

[0027] 4. Furthermore, in step 3, a multi-objective optimization task is designed by limiting the lithium plating overpotential and accelerating the charging speed. According to incremental MPC, its prediction domain is Y. p (k+1|k)=S x Δx(k)+y(k)+S u ΔI(k), S x S is the state gain matrix. u Here, ΔI(k) is the control gain matrix, and ΔI(k) is the control quantity, i.e., the change in charging current at each time step.

[0028] in:

[0029]

[0030] In the above formula, DA represents the output matrix D multiplied by the system matrix A, DB represents the output matrix D multiplied by the input matrix B, i represents the summation index, p represents the prediction domain size of the MPC method, and m represents the control domain size.

[0031] Finally, β represents the weight of the objective function, with a value ranging from 0 to 1. This weight can be adjusted according to actual needs. The final objective function is designed as follows:

[0032] J(ΔI(k))=||(1-β)Y p (k+1|k)|| 2 +||βΔI(k)|| 2

[0033] Through its Y p Both (k+1|k) and ΔI(k) will be limited to a certain range, and the minimum loss function value will be found within this range. The core idea is to find the optimal control current and reduce the fluctuation of charging current under the premise of limiting the overpotential range of lithium-ion plating.

[0034] Furthermore, in Step 4, an LSTM model is used to process the battery's current and voltage data to identify battery aging parameters, and the state-space equations in the model predictive control are updated based on these parameters. According to the battery aging mechanism, SEI layer growth and lithium plating are the main factors leading to battery aging. Therefore, the LSTM model is used to identify parameters in the battery charging current and voltage data to evaluate the key parameters characterizing battery aging, including the SEI layer film resistance R. sei and the volume ratio ε of solid-phase active materials in batteries s Based on these parameters, update the state-space equations, where the variable that needs to be modified is the SEI layer resistance R. sei and the specific surface area of ​​the material a s R represents the radius of the solid-state particle:

[0035] a s =3εs / R.

[0036] An electronic device, characterized in that it includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements a charging optimization method for suppressing lithium plating based on a mechanistic model as described in any one of the claims.

[0037] A non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the charging optimization method for suppressing lithium plating based on a mechanistic model as described in any one of the claims.

[0038] The advantages and beneficial effects of this invention are as follows:

[0039] This invention provides a lithium-ion battery charging strategy that considers battery aging based on model prediction and machine learning. The innovation lies in designing a model predictive control (MPC) strategy that combines the response characteristics of lithium plating potential to charging current, enabling real-time adjustment of the charging current. Furthermore, an LSTM network is used to periodically identify aging parameters to update the model in real time, preventing model distortion and ensuring that the charging strategy matches the actual battery performance, thereby improving charging efficiency. Compared to traditional methods, this method has the following advantages: traditional methods often use fixed charging strategies and cannot adjust the charging current in a timely manner to cope with changes in battery state; this method, through model predictive control, can respond to changes in battery state in real time. It can optimize among multiple objectives (such as balancing charging speed and battery life), ensuring that the charging objectives meet user needs. 。 By periodically identifying aging parameters using an LSTM network, the model can adapt to the battery aging process, thereby improving the accuracy and effectiveness of the charging strategy and avoiding algorithm failure due to ignoring battery aging. In summary, this method not only improves charging efficiency but also effectively adapts to battery aging, demonstrating a more advanced and scientific advantage than traditional methods. Attached Figure Description

[0040] Figure 1 The preferred embodiment of the present invention is shown in the technical roadmap of the present invention.

[0041] Figure 2 This is a flowchart for battery pack short circuit diagnosis.

[0042] Figure 3 A current diagram during the battery charging process.

[0043] Figure 4 A graph showing the change in lithium plating overpotential during battery charging.

[0044] Figure 5 To use LSTM for ε s The identification results.

[0045] Figure 6 To use LSTM for R sei The identification results. Detailed Implementation

[0046] The technical solutions of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0047] The technical solution of the present invention to solve the above-mentioned technical problems is:

[0048] A charging optimization method for suppressing lithium plating based on a mechanism model is characterized by the following steps:

[0049] Step 1: Simplify and linearize the battery electrochemical model;

[0050] Step 2: Based on the simplified battery electrochemical model, design the state space of the model predictive control method and obtain the response of the battery lithium plating overpotential to the charging current.

[0051] Step 3: By limiting the overpotential of lithium-ion battery plating and accelerating the charging speed, a multi-objective optimization task is designed, and the quadratic programming method is used to solve the battery charging current under the constraints.

[0052] Step 4: Using an LSTM model, identify battery aging parameters based on battery current and voltage data, and update the state-space equations in the model predictive control accordingly, enabling the charging strategy to dynamically adapt to the battery aging state.

[0053] In Step 2, a model predictive control method is constructed based on the battery electrochemical model. The state-space matrix related to the lithium-ion battery overpotential is as follows:

[0054] x(k+1)=Ax(k)+Bi(k)+C

[0055] y(k+1)=Dx(k+1)

[0056] Where x(k) is the value of the state space at time k, x = [C avg C surf JU ocp U sei η] T C avg C represents the average lithium-ion concentration in solid-phase particles. surf U represents the surface lithium-ion concentration of the solid particles, J represents the concentration flux of the lithium-ion intercalation reaction, and U represents the concentration flux of the lithium-ion intercalation reaction. ocp U represents the battery open-circuit voltage. seiLet η represent the voltage drop caused by the SEI layer of the battery, and η represent the overpotential of the lithium-ion battery intercalation reaction. i(k) represents the charging current at time k, i.e., the control target, and y(k) represents the lithium plating overpotential at time k. The matrix A is:

[0057]

[0058] Where R represents the radius of the solid-phase particle, and D eff C represents the solid-phase diffusion coefficient. max Let represent the maximum solid-phase lithium intercalation concentration, and 'a' be a first-order term in the fitting function relating the open-circuit potential of the electrode material to the amount of lithium intercalation. Its B matrix is:

[0059] Where A cell a represents the area of ​​the battery electrodes. s R represents the specific surface area of ​​the battery material, L represents the electrode length, F represents the Faraday constant, and R represents the electrode length. sei R represents the SEI layer film resistance of the battery. ct This represents the charge transfer impedance. Its C matrix is...

[0060] C = [0b 0 0 0 0] T

[0061] Where b is a constant term in the function relating the open-circuit potential of the electrode material to the amount of lithium intercalation, the D matrix is ​​represented as follows:

[0062] D = [0 0 0 1 1 1] T

[0063] In step 3, a multi-objective optimization task is designed by limiting the lithium plating overpotential and accelerating the charging speed. According to incremental MPC, its prediction domain is Y. p (k+1|k)=S x Δx(k)+y(k)+S u ΔI(k), where:

[0064]

[0065] Finally, β represents the weight of the loss function, which is designed as follows:

[0066] J(ΔI(k))=||(1-β)Y p (k+1|k)|| 2 +||βΔI(k)|| 2

[0067] Through its Y pBoth (k+1|k) and ΔI(k) will be constrained within a certain range, and the minimum loss function value will be sought within this range. The core idea is to find the optimal control current and reduce the fluctuation range of the charging current while limiting the overpotential range of lithium-ion battery plating.

[0068] In Step 4, an LSTM model is used to process the battery's current and voltage data to identify battery aging parameters. Based on these parameters, the state-space equations in the model's predictive control are estimated and updated. According to the battery aging mechanism, SEI layer growth and lithium plating are the main factors leading to battery aging. Therefore, by using an LSTM model to identify parameters from the battery charging current and voltage data, key parameters characterizing battery aging are evaluated, including the SEI layer film resistance R. sei and the volume ratio ε of solid-phase active materials in batteries s Based on these parameters, update the state-space equations, where the variable that needs to be modified is R. sei and a s :

[0069] a s =3ε s / R

[0070] This invention tests a simulation model of a lithium-ion battery by building a simulation experimental platform, and describes the specific implementation method of this invention in detail using simulation data. First, a P2D model is constructed using COMSOL; this paper uses the preset parameters of the battery electrochemical model within COMSOL software to construct the state-space equations for the model prediction method.

[0071] The implementation method of this invention uses the above-mentioned simulated battery as an example, and its implementation process includes the following steps:

[0072] Step 1: Construct the state-space equations for the model predictive control method based on the parameters of the battery electrochemical model.

[0073] Step 2: Based on the requirements for battery charging speed, avoiding capacity decay, and controlling current fluctuation amplitude in practical applications, the objective function and boundary conditions are set. This paper limits the lithium plating potential to between 0.1 and -0.1, and limits the current fluctuation at each time step to |ΔI(k)|<1A.

[0074] Step 3: Use a battery electrochemical aging coupled model to simulate charge-discharge cycle conditions, collecting current and voltage data for the last 500 seconds of each charge cycle, as well as the battery parameter R for each cycle. sei and ε s Generate a dataset of battery aging parameters.

[0075] Step 4: Using an LSTM model, learn and identify key parameter R through a battery aging dataset.sei and ε s like Figure 5 Figure 6 As shown.

[0076] Step 5: Enable the model prediction algorithm to control the charging current in real time, such as... Figure 3 As shown.

[0077] Step Six: After the battery has operated for a certain period of time, input the current and voltage data from the last 500 seconds of charging into the LSTM model, and adjust the parameter R in the battery electrochemical model. sei and ε s Parameter identification is performed, and the spatial state equation is updated. Through periodic updates, battery aging factors are incorporated into the charging strategy to avoid model distortion caused by battery aging. To account for computational resource consumption in practical applications, this paper sets parameter identification to be performed every 50 charge-discharge cycles.

[0078] In the experiment, by limiting the range of lithium plating potential and the fluctuation of current, a high charging speed of 6C-5C was obtained in the early stage of charging. At the same time, throughout the entire charging process, the lithium plating potential remained at around 0V, which can be regarded as no lithium plating phenomenon. Overall, high-speed charging with suppression of lithium plating reaction was achieved. In addition, the LSTM algorithm was successfully used to identify the aging parameters of the current battery, with the error controlled within 4%, ensuring that the entire charging strategy would not fail due to model distortion caused by battery aging.

[0079] It should be noted that for different batteries, parameter identification is required to obtain model parameters that are close to those of real batteries. For different batteries, the LSTM network needs to be retrained based on the battery model after parameter identification to estimate aging parameters.

[0080] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions.

[0081] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0082] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0083] The above embodiments should be understood as illustrative only and not as limiting the scope of protection of the present invention. After reading the description of the present invention, those skilled in the art can make various alterations or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A charge optimization method for suppressing lithium plating based on a mechanism model, characterized by, Includes the following steps: Step 1: Simplify the battery electrochemical model; Step 2: Based on the simplified battery electrochemical model, design the state space of the model predictive control method and obtain the response of the battery lithium plating overpotential to the charging current. Step 3: By limiting the overpotential of lithium-ion battery plating and accelerating the charging speed, a multi-objective optimization task is designed, and the quadratic programming method is used to solve the battery charging current under the constraints. Step 4: Using the LSTM model, identify battery aging parameters through battery current and voltage data, and update the state-space equations in the model predictive control accordingly, so that the charging strategy can dynamically adapt to the battery aging state. In Step 2, a model predictive control method is constructed based on the battery electrochemical model. The state-space matrix related to the lithium-ion battery overpotential is as follows: ； ； in It is the value of the state space at time k. ,in This represents the average lithium-ion concentration in solid-phase particles. This indicates the surface lithium ion concentration of solid particles. This represents the concentration flux of the lithium-ion intercalation reaction. Indicates the battery open-circuit voltage. This indicates the voltage drop caused by the SEI layer of the battery. This indicates the overpotential of the lithium-ion battery intercalation reaction. The target value is the charging current at time k. Let A represent the lithium plating overpotential at time k, where matrix A is: ； in Indicates the radius of a solid-phase particle. Represents the solid-phase diffusion coefficient. Indicates the maximum lithium intercalation concentration in the solid phase. It is a first-order term in the fitting function relating the open-circuit potential of the electrode material to the amount of lithium intercalation, and its B matrix is: ； in Indicates the area of ​​the battery electrodes. This indicates the specific surface area of ​​the battery material. Indicates electrode length, Denotes Faraday's constant. Indicates the SEI layer film resistance of the battery. Represents charge transfer impedance, and its C matrix is: ； in This is a constant term in the relationship between the open-circuit potential of the electrode material and the amount of lithium intercalation, represented by the D matrix as follows: 。 2. The charging optimization method for suppressing lithium plating based on a mechanism model according to claim 1, characterized in that, Step 1: Simplifying and linearizing the battery electrochemical model, specifically including: assuming the concentration distribution inside the particles is a parabolic function and combining it with the solid-phase concentration description of the single-particle model, the following can be derived: ； ； ； In the formula It is the flux through the orifice wall. This represents the average lithium-ion concentration in solid-phase particles. Represents the surface lithium ions of solid particles, where Indicates the radius of a solid-phase particle. Represents the solid-phase diffusion coefficient. This indicates the concentration flux of the lithium-ion intercalation reaction.

3. The charging optimization method for suppressing lithium plating based on a mechanism model according to claim 1, characterized in that, In step 3, a multi-objective optimization task is designed by limiting the lithium plating overpotential and accelerating the charging speed. Based on incremental MPC, its prediction domain is... , Here is the state gain matrix. For the control gain matrix, The control quantity is the change in charging current at each time step. in: ； ； In the above formula Represents the output matrix Multiply by the system matrix , Represents the output matrix Multiply by the input matrix , Indicates the summation index. This indicates the size of the prediction domain for the MPC method. Indicates the size of the control domain; at last The function weights in the objective function range from 0 to 1 and can be adjusted according to actual needs. The final objective function is designed as follows: + ； Through its and All of these will be limited to a certain range, and the minimum loss function value will be sought within this range. The core idea is to find the optimal control current and reduce the fluctuation range of charging current under the premise of limiting the overpotential range of lithium battery plating.

4. The charging optimization method for suppressing lithium plating based on a mechanism model according to claim 1, characterized in that: In Step 4, an LSTM model is used to process the battery's current and voltage data to identify battery aging parameters. Based on these parameters, the state-space equations in the model predictive control are estimated and updated. According to the battery aging mechanism, SEI layer growth and lithium plating are the main factors leading to battery aging. Therefore, the LSTM model is used to identify parameters in the battery charging current and voltage data to evaluate the key parameters characterizing battery aging, including SEI layer resistance. and the volume ratio of solid-phase active materials in batteries ; Based on these parameters, the state-space equations are updated, with the variable requiring modification being the SEI layer resistance. and material specific surface area , Represents the radius of a solid-phase particle: 。 5. An electronic device, characterized in that, The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the charging optimization method for suppressing lithium plating based on a mechanism model as described in any one of claims 1 to 4.

6. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the charging optimization method for suppressing lithium plating based on a mechanism model as described in any one of claims 1 to 4.

Images