Battery SOH evaluation method combining electrochemical model and machine learning

Abstract

The invention discloses a battery SOH evaluation method combining an electrochemical model and machine learning, and the method comprises the steps: constructing a target battery electrochemical simulation model, generating voltage curves and SOH data under the conditions of different temperatures, charging rates and different battery aging states, and constructing an SOH evaluation model and a voltage curve prediction model according to the voltage curves and the SOH data; a constant-current charging segment of the target battery in the current charging cycle is obtained in real time, the actually measured temperature and charging rate of the target battery corresponding to the constant-current charging segment are obtained, multiple SEI thicknesses and lithium precipitation thicknesses are preset, and multiple voltage curves are obtained through prediction of a voltage curve prediction model; acquiring a voltage curve segment closest to the initial SOC of the constant-current charging segment in the plurality of voltage curves, and calculating corresponding SEI thickness and lithium precipitation thickness; and obtaining an SOH estimated value according to the SOH evaluation model. According to the invention, accurate evaluation can be realized for a small data set and a novel battery.

Description

Technical Field

[0001] The present application relates to the technical field of battery SOH evaluation, and in particular to a battery SOH evaluation method and system. Background Art

[0002] As the global demand for clean energy increases, batteries have been widely used as an important energy storage device. However, batteries will experience capacity decay and increased internal resistance during long-term use, which directly affects their service life and safety. Therefore, accurate assessment of the battery state of health (SOH) has become an important means to ensure the safety and reliability of energy storage systems. Accurate SOH assessment can not only extend the battery life, but also identify potential risks in a timely manner and prevent situations that may lead to battery failure.

[0003] Battery SOH refers to the health of the battery's current state relative to its ideal state. SOH assessment includes the measurement and analysis of battery capacity, internal resistance, efficiency and other parameters to determine the remaining life and performance degradation of the battery. An accurate SOH assessment system should be able to monitor the battery's operating status in real time, predict future performance changes, and provide decision support on battery maintenance and replacement.

[0004] At present, the evaluation methods of battery SOH mainly include laboratory testing, physical models and data-driven statistical models. Although laboratory testing is highly accurate, it is costly and difficult to apply to real-time monitoring of actual projects; physical models are based on the electrochemical characteristics of batteries, but the parameters are complex and difficult to obtain accurately, making it difficult to model based on the electrochemical characteristics of batteries, and the practicality is not strong; data-driven statistical models are modeled through historical data and can be monitored in real time, but they are highly dependent on large data sets and difficult to handle small data sets and new batteries. Summary of the invention

[0005] Based on this, in response to the above technical problems, a battery SOH evaluation method combining electrochemical model and machine learning is provided to solve the problems of high cost of existing technology and difficulty in realizing SOH evaluation for small data sets and new batteries.

[0006] In a first aspect, a battery SOH evaluation method combining an electrochemical model and machine learning is provided, the method comprising:

[0007] Constructing an electrochemical simulation model of a target battery, and then generating voltage curves and SOH data under different temperatures, charging rates and different battery aging conditions based on the electrochemical simulation model, wherein the battery aging conditions include SEI thickness and lithium deposition thickness;

[0008] An SOH evaluation model is constructed based on the obtained SOH data under different aging conditions;

[0009] A voltage curve prediction model is constructed based on the voltage curves obtained under different temperatures, charging rates and different battery aging conditions;

[0010] Real-time acquisition of the constant current charging segment of the target battery in this charging cycle, acquisition of the actually measured temperature and charging rate of the target battery corresponding to the constant current charging segment, and preset multiple SEI thicknesses and lithium deposition thicknesses; multiple voltage curves are predicted by the trained voltage curve prediction model according to the actually measured temperature and charging rate and the preset multiple SEI thicknesses and lithium deposition thicknesses; acquisition of the voltage charging segment closest to the starting SOC of the constant current charging segment among the multiple voltage curves, and calculation of the corresponding SEI thickness and lithium deposition thickness, so as to obtain the actual SEI thickness and lithium deposition thickness of this charging cycle;

[0011] The actual SEI thickness and lithium plating thickness obtained in this charging cycle are input into the training SOH evaluation model to obtain the SOH estimation value.

[0012] In the above scheme, optionally, constructing an electrochemical simulation model of a target battery includes:

[0013] Construct an electrochemical simulation model of the target battery to describe the ion diffusion, electrode reaction, and potential change process inside the target battery;

[0014] The battery performance experimental data of the target battery at different temperatures and charging rates are obtained, thereby optimizing the electrochemical simulation model of the target battery.

[0015] In the above solution, further optionally, the SOH evaluation model adopts a Lasso regression model.

[0016] In the above solution, optionally, the voltage curve prediction model adopts a multi-output XGBoost model.

[0017] In the above solution, further optionally, during the training process of the voltage curve prediction model, the loss function is defined as the MSE of the output voltage point, and the MSE is calculated in the following way:

[0018]

[0019] where m is the number of output variables, n j is the number of samples of the jth output, y ij and are the true value and the predicted value respectively.

[0020] In the above solution, optionally, the preset multiple SEI thicknesses and lithium deposition thicknesses include:

[0021] Get the battery SEI thickness and lithium plating thickness of the last charging cycle, set the discrete steps of SEI thickness and lithium plating thickness, and get multiple preset SEI thickness and lithium plating thickness through the following sequence:

[0022]

[0023] Among them, S i-1 and L i-1 are the battery SEI thickness and lithium plating thickness of the previous charging cycle, ΔS and ΔL are the discrete step lengths of the set SEI thickness and lithium plating thickness, respectively, and n is 3 to 5.

[0024] In the above solution, optionally, obtaining the actual SEI thickness and lithium deposition thickness of this charging cycle includes:

[0025] Acquire multiple constant current charging segments of the target battery in this charging cycle in real time, and acquire the actual SEI thickness and lithium plating thickness for each constant current charging segment;

[0026] The actual SEI thickness and lithium deposition thickness of multiple constant current charged segments are averaged to obtain the actual SEI thickness and lithium deposition thickness of this charging cycle.

[0027] In a second aspect, a battery SOH evaluation system combining an electrochemical model and machine learning is provided, the system comprising:

[0028] Training data generation module: used to build an electrochemical simulation model of the target battery, and then generate voltage curves and SOH data under different temperatures, charging rates and different battery aging conditions based on the electrochemical simulation model. The battery aging conditions include SEI thickness and lithium deposition thickness;

[0029] SOH evaluation model building module: used to train the SOH evaluation model based on the obtained SOH data under different aging conditions;

[0030] Voltage curve prediction model building module: used to train the voltage curve prediction model based on the voltage curves obtained under different temperatures, charging rates and different battery aging conditions;

[0031] Actual SEI thickness and lithium deposition thickness acquisition module: used to obtain the constant current charging segment of the target battery in this charging cycle in real time, obtain the actually measured temperature and charging rate of the target battery corresponding to the constant current charging segment, and preset multiple SEI thicknesses and lithium deposition thicknesses; predict multiple voltage curves according to the actually measured temperature and charging rate and the preset multiple SEI thicknesses and lithium deposition thicknesses through the trained voltage curve prediction model; obtain the voltage charging segment closest to the starting SOC distance of the constant current charging segment in multiple voltage curves, and calculate the corresponding SEI thickness and lithium deposition thickness to obtain the actual SEI thickness and lithium deposition thickness of this charging cycle;

[0032] SOH estimation value acquisition module: used to input the actual SEI thickness and lithium plating thickness acquired in this charging cycle into the evaluation model for training SOH to obtain the SOH estimation value.

[0033] In a third aspect, a computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the battery SOH evaluation method combining electrochemical model and machine learning described in the first aspect are implemented.

[0034] In a fourth aspect, a computer program product includes a computer program / instruction, wherein when the computer program / instruction is executed by a processor, the steps of the battery SOH evaluation method combining an electrochemical model and machine learning as described in the first aspect are implemented.

[0035] This application has at least the following beneficial effects:

[0036] This application constructs an electrochemical simulation model of the target battery, predicts the voltage curve and SOH data under the same temperature, charging rate and different battery aging conditions according to the constructed electrochemical simulation model, and uses this as a training set to construct an SOH evaluation model and a voltage curve prediction model; then, by obtaining the temperature and charging rate actually measured in the constant current charging segment of the target battery in this charging cycle, and presetting a variety of SEI thicknesses and lithium plating thicknesses, multiple voltage curves are predicted by the trained voltage curve prediction model, and the voltage charging segment closest to the starting SOC of the constant current charging segment in multiple voltage curves is obtained, and the corresponding SEI thickness and lithium plating thickness are calculated to obtain the actual SEI thickness and lithium plating thickness of this charging cycle; the actual SEI thickness and lithium plating thickness of this charging cycle obtained are input into the training SOH evaluation model to obtain the SOH estimation value. Through the above method, accurate evaluation can also be achieved for small data sets and new batteries, and the cost is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1A schematic flow chart of a battery SOH evaluation method combining an electrochemical model and machine learning provided in one embodiment of the present application;

[0038] Figure 2 A detailed flowchart of a battery SOH evaluation method combining an electrochemical model and machine learning is provided for one embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0040] In one embodiment, Figure 1 As shown, a battery SOH evaluation method combining an electrochemical model and machine learning is provided, the method comprising:

[0041] Step S1: construct an electrochemical simulation model of the target battery, and then generate voltage curves and SOH data under different temperatures, charging rates and different battery aging conditions based on the electrochemical simulation model. The battery aging conditions include SEI thickness and lithium plating thickness.

[0042] Step S1 specifically includes the following contents:

[0043] Step S1.1: Construct an electrochemical simulation model of the battery to reflect the aging and attenuation mechanism of the battery. First, construct an electrochemical simulation model of the battery, and select, for example, the Pseudo-2D (P2D) model to reflect the internal electrochemical behavior of the battery. This model can accurately describe processes such as ion diffusion, electrode reaction, and potential change inside the battery. Aging mechanisms are added to the model, especially the growth of the SEI (solid electrolyte interface) film and the precipitation of lithium metal, to reflect the gradual attenuation of the battery's health state. The growth of the SEI film can be described by the following differential equation:

[0044]

[0045] Among them, L SEI represents SEI thickness, k SEI is the growth rate constant, I SEI is the relevant current density.

[0046] Step S1.2: Perform physical experimental tests on the battery to be evaluated to obtain test data of the battery under certain specific conditions. Test the battery performance data under different conditions through experiments. The experimental settings include different temperatures (such as 0℃, 25℃, etc.), different charging rates (such as 0.5C, 1C, 2C, etc.), and collect battery voltage, current, temperature and other data through constant current-constant voltage charge and discharge tests.

[0047] Step S1.3: Perform parameter identification on the electrochemical model based on the test data of the battery, and determine the parameters of the simulation model so that its simulation output results are close to the experimental data. Use the experimental data to perform parameter identification on the simulation model. The goal is to make the simulation results as close to the experimental data as possible by optimizing the parameters in the model. Usually, the least squares method or other optimization algorithms are used to minimize the mean square error (MSE) between the simulation output and the experimental data:

[0048]

[0049] Here, V sim and V exp are the voltage values ​​obtained from simulation and experiment, respectively.

[0050] Step 1.4: Generate training data sets for the battery under different temperatures, charge rates and different battery aging conditions based on the simulation model obtained in step 1.3. After the simulation model is determined, generate training data sets for the battery under different temperatures, charge rates, SEI thickness and lithium plating thickness conditions. Discretize the SEI thickness into X levels, the lithium plating thickness into Y levels, and the temperature and charge rate into M and N levels, respectively. Simulate the constant current charging process with SOC ranging from 0% to 100% under each combination of conditions (temperature, charge rate, SEI thickness, lithium plating thickness), and generate a total of M×N×X×Y data samples. Each sample contains the voltage curve under specific conditions and the corresponding SEI thickness, lithium plating thickness and SOH information.

[0051] Step S2: constructing an SOH evaluation model based on the obtained SOH data under different aging conditions.

[0052] The specific steps of step S2 include:

[0053] Step S2.1: Based on the battery simulation data generated in step 1, the Lasso regression method is used to train the SOH evaluation model. The model input is SEI thickness and lithium deposition thickness, and the output is the corresponding SOH. 1 Regularization), the input is SEI thickness and lithium deposition thickness, and the output is SOH state:

[0054]

[0055] Among them, x i is the input feature, and λ is the regularization parameter, which can effectively reduce feature redundancy and enhance the generalization ability of the model.

[0056] Step S3: constructing a voltage curve prediction model based on the obtained voltage curves under different temperatures, charging rates and different battery aging conditions.

[0057] The specific steps of step S3 include:

[0058] Step S3.1: Based on the battery simulation data generated in step S1, a multi-output XGBoost model is used to train the data model. In this step, we use a multi-output XGBoost model to train the data model. XGBoost is a decision tree model based on gradient boosting with high accuracy and performance. Multi-output XGBoost extends the functionality of XGBoost to support multiple output variables and predict multiple target values ​​at the same time.

[0059] In this specific application, the input features include temperature (T), rate (C), SEI thickness (L SEI ) and lithium deposition thickness (L Li ), which are the key factors for predicting the voltage curve. The output variable is 10 time-equidistant voltage points on the voltage curve (v 0 , v 1 , ..., v 10 ), and the time interval between each two voltage points (Δt). In the case of multiple outputs, XGBoost will establish a shared tree structure for all output variables, so that each tree considers the loss function of all outputs when modeling, thereby improving the overall prediction ability of the model.

[0060] In the specific XGBoost model training process, the loss function is defined as the MSE of the output voltage point, and the multi-output form can be expressed as:

[0061]

[0062] where m is the number of output variables, n j is the number of samples of the jth output, y ij and are the true value and the predicted value respectively. Multi-output XGBoost builds multiple base models by calculating the errors of all target values ​​in each iteration, thereby optimizing the overall performance.

[0063] The multi-output XGBoost model allows multiple voltage points (such as v 0 , v 1 , ..., v10 ), these outputs are critical for reconstructing the charging curve. In this way, we can simplify the complex voltage curve prediction problem to predicting the voltage value at 10 time points, and then complete the entire curve through methods such as linear interpolation.

[0064] Step S3.2: Using the predicted voltage point (v 0 , v 1 , ..., v 10 ) and Δt can be used to synthesize a complete voltage curve. Using these discrete voltage points combined with the time interval, a complete voltage curve is synthesized through interpolation methods (such as linear interpolation or high-order interpolation) to represent the charging behavior of the battery under specific conditions.

[0065] Step S4: Real-time acquisition of the constant current charging segment of the target battery in this charging cycle, acquisition of the actually measured temperature and charging rate of the target battery corresponding to the constant current charging segment, and preset multiple SEI thicknesses and lithium deposition thicknesses; prediction of multiple voltage curves based on the actually measured temperature and charging rate and the preset multiple SEI thicknesses and lithium deposition thicknesses by the trained voltage curve prediction model; acquisition of the voltage charging segment closest to the starting SOC of the constant current charging segment among the multiple voltage curves, and calculation of the corresponding SEI thickness and lithium deposition thickness, so as to obtain the actual SEI thickness and lithium deposition thickness of this charging cycle.

[0066] The specific steps of step S4 include:

[0067] Step S4.1: When evaluating the SOH of the i-th charging cycle online, a constant current charging segment in the charging cycle is intercepted. Assume that there are K constant current charging segments in the i-th charging cycle, and each segment corresponds to a different charging current.

[0068] Step S4.2: For the kth charging segment, use the multi-output XGBoost model in step 3, input the conditions of the constant current charging segment, and generate a voltage curve. Input charging rate C i，k and temperature T i，k , and the discrete SEI thickness and lithium deposition thickness The model output corresponds to (2n) 2 A complete voltage curve. In the process of discretizing SEI thickness and lithium deposition thickness, the parameter n determines the number of steps up and down from the current state in the discrete space. In the process of selecting n, it is necessary to comprehensively consider the balance between the efficiency and accuracy of the algorithm.

[0069] Algorithm efficiency: When n is large, the number of discrete levels increases, and the number of voltage curves generated in each iteration is (2n) 2As n increases, the amount of calculation increases quadratically, which significantly increases the computational cost and time complexity. Therefore, in order to maintain computational efficiency, the value of n should not be too large.

[0070] Model accuracy: On the other hand, choosing a smaller n may result in too large a discrete step size, which in turn leads to increased errors in estimating SEI thickness and lithium plating thickness. If the number of discrete points is insufficient, subtle changes in the voltage curve may not be captured, affecting the accuracy of the evaluation results. Therefore, it is necessary to ensure appropriate discrete accuracy.

[0071] Therefore, considering the balance between efficiency and accuracy, n is usually selected between 3 and 5. The value of n in this range can provide sufficient discrete density to capture the subtle changes in SEI thickness and lithium plating thickness, while avoiding excessive computational burden caused by too many discrete points. For example, when n = 3, we have (2n) 2 = 36 voltage curves need to be calculated, and the calculation cost is relatively low; when n = 5, there are (2n) 2 = 100 voltage curves, which requires a higher computational cost, but at the same time the estimation accuracy is also higher.

[0072] Selecting the range of n in this way can effectively balance the accuracy and computational complexity of voltage curve matching, so that the model can maintain reasonable accuracy without increasing the computational time too much, making the overall algorithm feasible in practical applications.

[0073] Step S4.3: The voltage curve generated by step S4.2 is intercepted so that it has the same starting SOC as the kth charging segment. From the intercepted generated voltage curve segments, find the one with the closest Euclidean distance to the kth charging segment. Match the generated voltage curve segment with the kth real segment, and use the Euclidean distance to measure the similarity.

[0074] Euclidean distance is a common metric used to calculate the distance between two vectors. For voltage curve matching, we can regard the generated voltage curve segments and the real charging segments as vectors, and set the sets of voltage values ​​at several time points as the components of these vectors. Assume that the charging segment voltage data is v = [v 1 , v 2 , ..., v n ], the generated voltage curve segment is u=[u 1 ,u 2 , ..., u n ], then the Euclidean distance between them can be defined as:

[0075]

[0076] Here, v i and ui They represent the voltage values ​​of the real charging segment and the generated voltage curve segment at the i-th time point. By calculating the Euclidean distance d(v, u), the similarity of the two voltage curves in the time dimension can be determined.

[0077] Select the voltage curve with the minimum Euclidean distance: For the generated (2n) 2 The generated curve with the smallest Euclidean distance is found, and the SEI thickness and lithium deposition thickness corresponding to it are recorded as S i，k and L i，k This approach ensures that the voltage profile that is most similar to the actual charging segment is selected, thereby improving the accuracy of SEI thickness and lithium plating thickness estimation.

[0078] Step S4.4: For all constant current charging segments of the i-th charging cycle, the corresponding SEI thickness and lithium deposition thickness are evaluated according to the methods of step 4.2 and step 4.3, and the average values ​​of the corresponding SEI thickness and lithium deposition thickness are calculated to obtain:

[0079]

[0080] This method can reduce the error of single fragment matching and improve the accuracy of the overall estimation.

[0081] Step S5: Input the actual SEI thickness and lithium plating thickness obtained in this charging cycle into the evaluation model for training SOH to obtain the SOH estimation value.

[0082] Step S5.1: Apply the SOH evaluation model obtained in step 2 and input the SEI thickness S obtained in step 4 i and lithium deposition thickness L i , calculate the SOH of the ith cycle i . i and L i Input into the Lasso regression model to obtain the estimated SOH value of the i-th cycle:

[0083] SOH i =f(S i , L i )

[0084] In the above-mentioned battery SOH evaluation method, an electrochemical simulation model of the target battery is constructed, and the voltage curve and SOH data under the same temperature, charging rate and different battery aging conditions are predicted according to the constructed electrochemical simulation model, and this is used as a training set to construct an SOH evaluation model and a voltage curve prediction model; then, the temperature and charging rate actually measured in the constant current charging segment of the target battery in this charging cycle are obtained, and multiple SEI thicknesses and lithium plating thicknesses are preset, and multiple voltage curves are predicted by the trained voltage curve prediction model, and the voltage charging segment closest to the starting SOC of the constant current charging segment in the multiple voltage curves is obtained, and the corresponding SEI thickness and lithium plating thickness are calculated to obtain the actual SEI thickness and lithium plating thickness of this charging cycle; the actual SEI thickness and lithium plating thickness of this charging cycle are input into the training SOH evaluation model to obtain the SOH estimation value. Through the above method, accurate evaluation can also be achieved for small data sets and new batteries, and the cost is low.

[0085] In one embodiment, a battery SOH evaluation system combining an electrochemical model and machine learning is provided, wherein the system comprises:

[0086] Training data generation module: used to build an electrochemical simulation model of the target battery, and then generate voltage curves and SOH data under different temperatures, charging rates and different battery aging conditions based on the electrochemical simulation model. The battery aging conditions include SEI thickness and lithium deposition thickness;

[0087] SOH evaluation model building module: used to train the SOH evaluation model based on the obtained SOH data under different aging conditions;

[0088] Voltage curve prediction model building module: used to train the voltage curve prediction model based on the voltage curves obtained under different temperatures, charging rates and different battery aging conditions;

[0089] Actual SEI thickness and lithium deposition thickness acquisition module: used to obtain the constant current charging segment of the target battery in this charging cycle in real time, obtain the actually measured temperature and charging rate of the target battery corresponding to the constant current charging segment, and preset multiple SEI thicknesses and lithium deposition thicknesses; predict multiple voltage curves according to the actually measured temperature and charging rate and the preset multiple SEI thicknesses and lithium deposition thicknesses through the trained voltage curve prediction model; obtain the voltage charging segment closest to the starting SOC distance of the constant current charging segment in multiple voltage curves, and calculate the corresponding SEI thickness and lithium deposition thickness to obtain the actual SEI thickness and lithium deposition thickness of this charging cycle;

[0090] SOH estimation value acquisition module: used to input the actual SEI thickness and lithium plating thickness acquired in this charging cycle into the evaluation model for training SOH to obtain the SOH estimation value.

[0091] For the specific definition of a battery SOH evaluation system that combines electrochemical models with machine learning, please refer to the definition of a battery SOH evaluation method that combines electrochemical models with machine learning above, which will not be repeated here. Each module in the above-mentioned battery SOH evaluation system that combines electrochemical models with machine learning can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0092] In one embodiment, a computer device is provided, which may be a server. The computer device includes a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above-mentioned battery SOH evaluation method combining an electrochemical model with machine learning is implemented.

[0093] In one embodiment, a computer program product is also provided, including computer program I instructions, which, when executed by a processor, involve all or part of the processes in the above-mentioned embodiment method.

[0094] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0095] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0096] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A battery SOH evaluation method combining electrochemical model and machine learning, characterized in that: The method comprises: Constructing an electrochemical simulation model of a target battery, and then generating voltage curves and SOH data under different temperatures, charging rates and different battery aging conditions based on the electrochemical simulation model, wherein the battery aging conditions include SEI thickness and lithium deposition thickness; An SOH evaluation model is constructed based on the obtained SOH data under different aging conditions; A voltage curve prediction model is constructed based on the voltage curves obtained under different temperatures, charging rates and different battery aging conditions; Real-time acquisition of the constant current charging segment of the target battery in this charging cycle, acquisition of the actually measured temperature and charging rate of the target battery corresponding to the constant current charging segment, and preset multiple SEI thicknesses and lithium deposition thicknesses; multiple voltage curves are predicted by the trained voltage curve prediction model according to the actually measured temperature and charging rate and the preset multiple SEI thicknesses and lithium deposition thicknesses; acquisition of the voltage charging segment closest to the starting SOC of the constant current charging segment among the multiple voltage curves, and calculation of the corresponding SEI thickness and lithium deposition thickness, so as to obtain the actual SEI thickness and lithium deposition thickness of this charging cycle; The actual SEI thickness and lithium plating thickness obtained in this charging cycle are input into the training SOH evaluation model to obtain the SOH estimation value.

2. The battery SOH evaluation method combining electrochemical model and machine learning according to claim 1 is characterized in that: The electrochemical simulation model of the target battery is constructed as follows: Construct an electrochemical simulation model of the target battery to describe the ion diffusion, electrode reaction, and potential change process inside the target battery; The battery performance experimental data of the target battery at different temperatures and charging rates are obtained, thereby optimizing the electrochemical simulation model of the target battery.

3. The battery SOH evaluation method combining electrochemical model and machine learning according to claim 2 is characterized in that: The SOH evaluation model adopts the Lasso regression model.

4. The battery SOH evaluation method combining electrochemical model and machine learning according to claim 1, characterized in that: The voltage curve prediction model adopts a multi-output XGBoost model.

5. The battery SOH evaluation method combining electrochemical model and machine learning according to claim 4 is characterized in that: During the training process of the voltage curve prediction model, the loss function is defined as the MSE of the output voltage point, and the MSE is calculated in the following way: where m is the number of output variables, n j is the number of samples of the jth output, y ij and are the true value and the predicted value respectively.

6. The battery SOH evaluation method combining electrochemical model and machine learning according to claim 1, characterized in that: The preset multiple SEI thicknesses and lithium plating thicknesses include: Get the battery SEI thickness and lithium plating thickness of the last charging cycle, set the discrete steps of SEI thickness and lithium plating thickness, and get multiple preset SEI thickness and lithium plating thickness through the following sequence: Among them, S i-1 and L i-1 are the battery SEI thickness and lithium plating thickness of the previous charging cycle, ΔS and ΔL are the discrete step lengths of the set SEI thickness and lithium plating thickness, respectively, and n is 3 to 5.

7. The battery SOH evaluation method combining electrochemical model and machine learning according to claim 1, characterized in that: The method of obtaining the actual SEI thickness and lithium plating thickness of the current charging cycle includes: Acquire multiple constant current charging segments of the target battery in this charging cycle in real time, and acquire the actual SEI thickness and lithium plating thickness for each constant current charging segment; The actual SEI thickness and lithium deposition thickness of multiple constant current charged segments are averaged to obtain the actual SEI thickness and lithium deposition thickness of this charging cycle.

8. A battery SOH evaluation system combining electrochemical model and machine learning, characterized in that: The system comprises: Training data generation module: used to build an electrochemical simulation model of the target battery, and then generate voltage curves and SOH data under different temperatures, charging rates and different battery aging conditions based on the electrochemical simulation model. The battery aging conditions include SEI thickness and lithium deposition thickness; SOH evaluation model building module: used to train the SOH evaluation model based on the obtained SOH data under different aging conditions; Voltage curve prediction model building module: used to train the voltage curve prediction model based on the voltage curves obtained under different temperatures, charging rates and different battery aging conditions; Actual SEI thickness and lithium deposition thickness acquisition module: used to obtain the constant current charging segment of the target battery in this charging cycle in real time, obtain the actually measured temperature and charging rate of the target battery corresponding to the constant current charging segment, and preset multiple SEI thicknesses and lithium deposition thicknesses; predict multiple voltage curves according to the actually measured temperature and charging rate and the preset multiple SEI thicknesses and lithium deposition thicknesses through the trained voltage curve prediction model; obtain the voltage charging segment closest to the starting SOC distance of the constant current charging segment in multiple voltage curves, and calculate the corresponding SEI thickness and lithium deposition thickness to obtain the actual SEI thickness and lithium deposition thickness of this charging cycle; SOH estimation value acquisition module: used to input the actual SEI thickness and lithium plating thickness acquired in this charging cycle into the evaluation model for training SOH to obtain the SOH estimation value.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images