Battery SOH estimation method based on constant-voltage charging current curve

Abstract

The invention discloses a battery SOH estimation method based on a constant-voltage charging current curve, and the method comprises the steps: extracting a health factor HI1, a health factor HI2 and a health factor HI3 which reflect the aging data of a battery based on each round of the constant-voltage charging current curve in the whole-cycle aging process of the battery, and carrying out the normalization processing of the health factor HI1, the health factor HI2 and the health factor HI3; according to the extracted normalized values of the health factor HI1, the health factor HI2 and the health factor HI3, obtaining related parameters based on an optimization algorithm, and training a data model; and inputting the normalized values of the health factors HI1, HI2 and HI3 into the training data model, and calculating the SOH of the battery. According to the current curve based on the constant voltage charging stage disclosed by the invention, the health factors reflecting the aging information of the battery are acquired, the health factors with higher applicability and mechanism property are extracted, and an optimization algorithm and a data model are constructed, so that the health state of the battery can be accurately estimated, and a battery health state result with higher precision can be obtained.

Description

Technical Field

[0001] The present invention belongs to the field of health status estimation of energy storage batteries, and more specifically, relates to a method for extracting battery aging information based on a constant voltage charging current curve, and estimating the health status of energy storage batteries using an optimization algorithm and a data model. Background Art

[0002] Energy storage batteries are widely used in various portable devices, electric vehicles and energy storage systems. The state of health (SOH) of energy storage batteries is one of its important status parameters and can be used as a key indicator to evaluate battery performance and life. A decrease in SOH will lead to a decrease in battery discharge efficiency, an increase in internal resistance, and even safety accidents such as fire. Therefore, accurately estimating the SOH of energy storage batteries can effectively optimize the use and maintenance strategies of energy storage batteries, reduce replacement costs and environmental impacts, and ensure efficient and stable use of batteries.

[0003] Energy storage battery state estimation methods are mainly divided into model-based methods and data-driven methods. Traditional model-based methods mainly obtain battery aging characteristic parameters through empirical models, electrochemical models and equivalent circuit models, and estimate the battery SOH by combining Kalman filtering, particle filtering and other methods. The advantages of model-based health state estimation methods are strong physical interpretation and low data requirements, but their modeling is complex, the computational cost is high, and the model parameters are difficult to obtain. The adaptability to actual working conditions is limited, which may lead to increased estimation errors in dynamic or extreme environments.

[0004] In addition, the existing data-driven methods are mainly divided into the extraction of the health indicator (HI) of the energy storage battery and the construction of the SOH estimation model. By extracting parameters that can characterize battery aging from the physical external characteristic parameters of the energy storage battery, such as voltage, current, and temperature, and then using artificial intelligence algorithms such as support vector machines, Gaussian process regression, and artificial neural networks to obtain the mapping relationship between aging parameters and SOH, the battery's SOH is estimated. The data-driven method does not require a physical model, has strong adaptability, good real-time performance, and can mine potential patterns from data. It is currently the most promising SOH estimation method for actual production.

[0005] However, current data-driven methods extract the health factor HI by analyzing the characteristics of the physical parameters of the entire battery charging and discharging process. However, the existing health factor HI extraction methods have certain limitations. For example, due to the differences in users' battery usage habits, it is not universal to extract the health factor HI only from the battery discharge phase. Secondly, the difference in battery discharge cut-off voltage and capacity will also lead to differences in the starting state of charge (SOC) of the next charging cycle. Therefore, the method of extracting the health factor HI by analyzing the constant current phase or the complete charging process data is also limited. Summary of the invention

[0006] In response to the defects of the prior art and the need for improvement, the present invention discloses a method for collecting health factors reflecting battery aging information based on the current curve in the constant voltage charging stage, extracting health factors with stronger applicability and mechanistic rationality, and constructing an optimization algorithm and data model to estimate the health status of the energy storage battery and obtain more accurate battery health status results.

[0007] The battery SOH estimation method based on the constant voltage charging current curve disclosed in the present invention includes:

[0008] Step S1, during the full cycle aging process of the battery, based on each round of the constant voltage charging current curve, extract the health factor HI1, health factor HI2 and health factor HI3 reflecting the battery aging data and normalize them respectively;

[0009] Step S2, according to the normalized values ​​of the extracted health factor HI1, the health factor HI2 and the health factor HI3 data, obtaining relevant parameters based on an optimization algorithm, and training a data model; and

[0010] Step S3, inputting the normalized values ​​of the health factors HI1, HI2 and HI3 into the training data model to calculate the SOH of the battery.

[0011] The step S1 specifically includes:

[0012] Step S11: pre-processing the capacity data of the battery to obtain the SOH data of the battery:

[0013]

[0014] Among them, Q n is the current capacity of the battery, Q o is the rated capacity of the battery;

[0015] Step S12: extracting the health factors HI1, HI2 and HI3, wherein the health factor HI1 is the time taken for the current to change from a preset current to a target current during the constant voltage charging stage, the health factor HI2 is the charging current turning point information, and the health factor HI3 is the radius of the charging current curve fitting circle; and

[0016] Step S13: Normalize the health factors HI1, HI2 and HI3:

[0017]

[0018] Among them, x' i The normalized values ​​of the health factors HI1, HI2 and HI3 in the ith cycle, x i are the values ​​of the health factors HI1, HI2 and HI3 in the ith cycle, and min(X) and max(X) are the minimum and maximum values ​​of the health factors HI1, HI2 and HI3 respectively.

[0019] The step S2 specifically includes:

[0020] Step S21: construct the training data model as a multi-core extreme learning machine model, and introduce the kernel method:

[0021] f(x')=Hβ

[0022] Among them, x' is the input sample data, f(x') is the output result, H is the feature mapping matrix of the hidden layer, and β is the weight between the hidden layer and the output layer. Its expression is:

[0023]

[0024] Among them, I is the identity matrix, c is the regularization parameter, T is the target vector of the training sample data, and the kernel function matrix model is:

[0025] H*=HH T

[0026]

[0027] Among them, x i ' is a vector consisting of the normalized values ​​of the health factors HI1, HI2 and HI3 of the input feature under the i-th cycle, x' j is the vector composed of the normalized values ​​of the health factors HI1, HI2 and HI3 of the input feature under the j-th cycle, H* is the similarity matrix, and the output result is:

[0028]

[0029] Wherein, x' is a vector composed of the normalized values ​​of the health factors HI1, HI2 and HI3 of the input feature under a specific cycle, x' n A vector consisting of the normalized values ​​of the health factors HI1, HI2 and HI3 of the input feature in the nth cycle.

[0030] Step S22: Optimize using the optimization algorithm to obtain the optimal solution of the relevant parameters required in the multi-core extreme learning machine model.

[0031] Preferably, step S21 further includes combining the Poly kernel function and the RBF kernel function to obtain a hybrid kernel function:

[0032] K m =ωK RBF +(1-ω)K Poly

[0033] Where ω is the weight coefficient,

[0034] K Poly (x',x i ')=(x'·x i '+b) d

[0035]

[0036] Wherein, x' is a vector composed of the normalized values ​​of the health factors HI1, HI2 and HI3 of the input feature in a specific cycle, and x i ' is a vector composed of the normalized values ​​of the health factors HI1, HI2 and HI3 of the input feature under the i-th cycle, b, d are the parameters of the Poly kernel function, and σ is the parameter of the RBF kernel function.

[0037] Optionally, the optimization algorithm in step S22 is an Osprey optimization algorithm, which specifically includes:

[0038] Step S31: Set the number of ospreys T, the maximum number of iterations N, and the parameter range to be optimized;

[0039] Step S32: Initialize the relevant parameters c, b, d, σ, ω;

[0040] Step S33: Evaluate the MSE of the fitness SOH training result;

[0041] Step S34: updating the relevant parameters c, b, d, σ, ω;

[0042] Step S35: Boundary check; and

[0043] Step S36: Obtain the optimal solution of the relevant parameters c, b, d, σ, ω.

[0044] Optionally, step S3 specifically includes:

[0045] Step S41: input the normalized values ​​of the health factors HI1, HI2 and HI3;

[0046] Step S42: Obtaining the optimal solution of the relevant parameters calculated by the optimization algorithm;

[0047] Step S43: training the data model;

[0048] Step S44: testing the data model; and

[0049] Step S45: Output the SOH estimation result of the battery.

[0050] Optionally, the optimization algorithms include: Osprey optimization algorithm, particle swarm optimization algorithm, grey wolf optimization algorithm and whale optimization algorithm.

[0051] Optionally, the training data model includes: a relevance vector machine model, a support vector machine model, a Gaussian process regression model and a multi-core extreme learning machine model.

[0052] The present invention discloses a method for collecting health factors reflecting battery aging information based on a current curve in a constant voltage charging stage, extracting health factors with stronger applicability and mechanistic rationality, and constructing an optimization algorithm and a data model to estimate the health status of an energy storage battery and obtain a battery health status result with higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1a It is a schematic diagram of a battery health factor HI1 and a battery health factor HI2 based on a constant voltage charging current curve according to an embodiment of the present invention;

[0054] Figure 1b A schematic diagram of a battery health factor HI3 based on a constant voltage charging current curve according to an embodiment of the present invention;

[0055] Figure 2 This is a flow chart of a method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention;

[0056] Figure 3 In the flowchart of a method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention, Figure 2 Detailed step diagram of step S1;

[0057] Figure 4 In the flowchart of a method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention, Figure 2Detailed step diagram of step S2;

[0058] Figure 5 In the flowchart of the method for estimating battery SOH based on a constant voltage charging current curve according to one embodiment of the present invention, when the optimization algorithm is the Osprey optimization algorithm and the training data model is the multi-core extreme learning machine model, the flowchart of estimating battery SOH is shown;

[0059] Figure 6 A method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention, an iterative process diagram of tuning a multi-core extreme learning machine model using an Osprey optimization algorithm; and

[0060] Figure 7 A method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention, and a comparison chart between an actual battery SOH value and an estimated battery SOH value. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention 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 invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0062] Figure 1a FIG. 1 is a schematic diagram of a battery health factor HI1 and a battery health factor HI2 based on a constant voltage charging current curve according to an embodiment of the present invention. Figure 1a As shown, taking the battery constant voltage charging current curve under the condition of battery cycle 1 and cycle 161 as an example, as the number of cycles increases and the battery ages, the dynamic process inside the battery slows down, resulting in a gradual decrease in the current decay rate in the constant voltage charging stage, and a prolonged constant voltage charging duration, which can characterize the battery aging process. Considering the possibility of incomplete charging, the middle part of the current is selected for experiment. In one embodiment of the present invention, during the full cycle aging process of the battery, the health factor HI1 is the time taken for the charging current to go from the preset current to the target current under the constant voltage charging environment, for example, the preset current is 1.5A, the target current is 0.5A, then the time interval from the charging current 1.5A to the target current 0.5A is the health factor HI1; in addition, the starting point and the end point of the charging current curve are connected to obtain the line segment L, for example, the starting point is the preset current point, the end point is the target current point, and the distance from the point on the charging current curve to the line segment L is calculated, where the time corresponding to the maximum distance point is the health factor HI2, also known as the charging current turning point information. It should be understood by those skilled in the art that the values ​​of the preset current and the target current here are only examples of the present invention and cannot be used as a limitation of the present invention.

[0063] Figure 1b This is a schematic diagram of a battery health factor HI3 based on a constant voltage charging current curve according to an embodiment of the present invention. As shown in the figure, taking the battery constant voltage charging current curves of battery cycle 1 and cycle 161 as examples, the charging current curve is fitted into a circle using a polynomial fitting method, and the radius of the circle is taken as the health factor HI3. As can be seen from the figure, the health factor HI3 gradually increases as the battery ages.

[0064] Figure 2 This is a flow chart of a method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention. Specifically, the following steps are included: Step S1: During the full cycle aging process of the battery, based on each round of the constant voltage charging current curve, extract the health factor HI1, health factor HI2 and health factor HI3 reflecting the battery aging data and normalize them; Step S2, according to the extracted normalized data of health factor HI1, health factor HI2 and health factor HI3, obtain relevant parameters and training data model based on the optimization algorithm; and Step S3, input the normalized values ​​of health factors HI1, HI2 and HI3 into the training data model to calculate the battery SOH.

[0065] Specifically, health factors HI1, HI2 and HI3 are data reflecting battery aging information extracted based on each round of constant voltage charging current curve during the full cycle aging of the battery. The optimization algorithms here include but are not limited to: Osprey optimization algorithm, particle swarm optimization algorithm, gray wolf optimization algorithm and whale optimization algorithm; training data models include but are not limited to: relevance vector machine model, support vector machine model, Gaussian process regression model and multi-core extreme learning machine model. Those skilled in the art should understand that the batteries here include but are not limited to lithium batteries and liquid metal batteries.

[0066] Figure 3 This is a flow chart of a method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention. Figure 2 The detailed step diagram of step S1 in FIG. 1 specifically includes the following steps:

[0067] Step S11: pre-process the battery capacity data to obtain the battery SOH data:

[0068]

[0069] Among them, Q n is the current capacity of the battery, Q o is the rated capacity of the battery.

[0070] Step S12: extracting health factors HI1, HI2 and HI3, wherein the health factor HI1 is the time taken for the current to change from a preset current to a target current during the constant voltage charging stage, the health factor HI2 is the charging current turning point information, and the health factor HI3 is the radius of the charging current curve fitting circle; and

[0071] Step S13: Normalize the health factors HI1, HI2 and HI3:

[0072]

[0073] Among them, x' i The normalized values ​​of health factors HI1, HI2 and HI3 in the ith cycle, x i are the values ​​of health factors HI1, HI2 and HI3 in the ith cycle, min(X) and max(X) are the minimum and maximum values ​​of health factors HI1, HI2 and HI3 respectively.

[0074] Preferably, the health factors HI1, HI2, HI3 and the battery SOH are subjected to correlation analysis, and it can be verified that the health factors HI1, HI2, HI3 and the battery SOH have a strong correlation, indicating that the health factors selected by the present invention can accurately reflect the battery aging. Correlation analysis is a method well known to those skilled in the art, and will not be described here to avoid redundancy.

[0075] Figure 4 In the flowchart of a method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention, Figure 2 The detailed step diagram of step S2 in FIG. 1 specifically includes the following steps:

[0076] Step S21: construct the training data model as a multi-core extreme learning machine model, and introduce the kernel method:

[0077] f(x')=Hβ

[0078] Among them, x' is the input sample data, f(x') is the output result, H is the feature mapping matrix of the hidden layer, and β is the weight between the hidden layer and the output layer. Its expression is:

[0079]

[0080] Among them, I is the identity matrix, c is the regularization parameter, T is the target vector of the training sample data, and the kernel function matrix model is:

[0081] H*=HH T

[0082]

[0083] Among them, x i ' is a vector composed of the normalized values ​​of the input characteristic health factors HI1, HI2 and HI3 under the i-th cycle, x' j is the vector composed of the normalized values ​​of the input characteristic health factors HI1, HI2 and HI3 in the jth cycle, H* is the similarity matrix, and the output result is:

[0084]

[0085] Among them, x' is a vector composed of the normalized values ​​of the input characteristic health factors HI1, HI2 and HI3 in a specific cycle, x' n It is a vector composed of the normalized values ​​of the input features HI1, HI2 and HI3 in the nth cycle.

[0086] Specifically, the Extreme Learning Machine (ELM) is a fast learning algorithm for a single hidden layer feedforward neural network, which has significant advantages in processing large-scale data, but its randomly initialized hidden layer parameters may cause model instability and is only applicable to some nonlinear problems. The present invention is illustrated by taking a multi-core extreme learning machine as an example, which cannot be used as a limitation of the present invention.

[0087] Step S22: Optimize using an optimization algorithm to obtain the optimal solution for the relevant parameters required in the multi-core extreme learning machine model.

[0088] In one embodiment, the osprey optimization algorithm is selected to tune the multi-core extreme learning machine model to obtain the optimal solution of the relevant parameters. The osprey optimization algorithm efficiently solves complex optimization problems by simulating the process of osprey circling in the air to find prey and diving quickly to catch prey. The osprey optimization algorithm combines the advantages of global search and local search, and is superior to traditional optimization algorithms in convergence speed and solution accuracy. The global search process aims to explore the entire space to avoid falling into the local optimum, thereby increasing the possibility of finding the global optimal solution. The local search process further optimizes the solution based on the global search, fine-tunes the individual positions of the osprey to improve the accuracy of the solution, and speeds up the convergence speed, so as to find a better solution to the problem.

[0089] Preferably, in one embodiment, Figure 4 Step S21 in the embodiment also includes combining the Poly kernel function and the RBF kernel function to obtain a hybrid kernel function. The Poly kernel function characterizes the overall attenuation trend of the battery SOH, and the Poly kernel function characterizes the phenomenon of local fluctuation of the SOH. By combining the two, a multi-core extreme learning machine model is constructed to improve the accuracy of battery SOH estimation. The Poly kernel function and the RBF kernel function are respectively:

[0090] KPoly (x',x i ')=(x'·x i '+b) d

[0091]

[0092] Among them, x' is a vector composed of the normalized values ​​of the input characteristic health factors HI1, HI2 and HI3 in a specific cycle, x i ' is a vector composed of the normalized values ​​of the input characteristic health factors HI1, HI2 and HI3 in the i-th cycle, b and d are the parameters of the Poly kernel function, and σ is the parameter of the RBF kernel function. The hybrid kernel function is obtained based on the Poly kernel function and the RBF kernel function as follows:

[0093] K m =ωK RBF +(1-ω)K Poly

[0094] The mixed kernel function K m is a linear combination of the Poly kernel function and the RBF kernel function, and ω is a weight coefficient. In one embodiment of the present invention, the hybrid kernel function K m Substituting into the output equation f(x'), the output result is more accurate in estimating the battery SOH than the single kernel function.

[0095] Figure 5 In the flowchart of the method for estimating battery SOH based on the constant voltage charging current curve according to one embodiment of the present invention, when the optimization algorithm is the Osprey optimization algorithm and the training data model is the multi-core extreme learning machine model, the flowchart of estimating battery SOH is shown. Figure 5 The dashed box on the left is the Osprey optimization algorithm flow chart, and the dashed box on the right is the multi-core extreme learning machine model training flow chart. Figure 5 As shown, the Osprey optimization algorithm includes the following steps:

[0096] Step S31: Set the number of ospreys T, the maximum number of iterations N, and the parameter range to be optimized;

[0097] Step S32: Initialize relevant parameters c, b, d, σ, ω;

[0098] Step S33: Evaluate the MSE of the fitness SOH training result;

[0099] Step S34: Update relevant parameters c, b, d, σ, ω;

[0100] Step S35: Boundary check; and

[0101] Step S36: Obtain the optimal solution of relevant parameters c, b, d, σ, ω.

[0102] The multi-core extreme learning machine model training and calculation process includes the following steps:

[0103] Step S41: input the normalized values ​​of health factors HI1, HI2 and HI3;

[0104] Step S42: Obtaining the optimal solution of the relevant parameters calculated by the optimization algorithm;

[0105] Step S43: training data model;

[0106] Step S44: testing the multi-core extreme learning machine model; and

[0107] Step S45: Output the battery SOH estimation result.

[0108] Specifically, the battery health factors HI1, HI2 and HI3 are normalized and their corresponding battery SOH are divided into training set, validation set and test set respectively. In one embodiment of the present invention, the ratios of the training set, validation set and test set are set to 48%, 12% and 40% respectively, and experimental verification is performed. Those skilled in the art should understand that the ratio setting here is only for illustrating the present invention and cannot be regarded as limiting the present invention.

[0109] First, the parameters of the osprey optimization algorithm are initialized. In the present invention, the maximum number of iterations T of the osprey optimization algorithm is set to 50 times, and the number of osprey individuals N is 50; the search ranges of the parameters to be optimized b, σ, ω, and c are all set to [0.01, 100], and the search range of d is set to [-2, 2]. Each osprey individual represents a set of parameter combinations, that is, there will be 50 sets of parameter combinations in one cycle, among which there is a set of the best parameters; then the 50 sets of parameters enter the next iteration to find a better parameter combination, and there will be another 50 sets of solutions. After 50 iterations, a total of 2500 sets of solutions are generated, and the best combination of the last 50 is selected as the optimization result of the parameters b, σ, ω, c, and d.

[0110] Secondly, a fitness function is established. In one embodiment of the present invention, the mean square error (MSE) is used as the fitness function. The specific expression is as follows:

[0111]

[0112] Among them, y est and real are the estimated value and true value of battery SOH respectively, and n is the number of cycles.

[0113] Then, the Osprey optimization algorithm is used to optimize the parameters of the multi-core extreme learning machine model. The model is trained and the fitness value of each time on the validation set is recorded until the maximum number of iterations T is reached. The training is stopped and the optimal combination of relevant parameters c, b, d, σ, and ω is recorded.

[0114] Finally, the model is trained based on the optimal combination of relevant parameters c, b, d, σ, and ω, and the battery SOH of the test set is estimated to verify the model performance.

[0115] Those skilled in the art should understand that the above-mentioned Osprey optimization algorithm and multi-core extreme learning machine model are only one embodiment of the present invention, the purpose of which is to illustrate the present invention and verify the effect of the present invention, and the concept of the present invention is not limited to the above embodiment.

[0116] Figure 6 A method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention, and an iterative process diagram of tuning a multi-core extreme learning machine model using the Osprey optimization algorithm. As shown in the figure, the iterative error uses the root mean square error, the error keeps decreasing, and finally converges to obtain the optimal combination of related parameters.

[0117] Figure 7 A method for estimating battery SOH based on a constant voltage charging current curve according to an embodiment of the present invention, and a comparison chart of the actual battery SOH value and the estimated battery SOH value. As can be seen from the figure, under the hybrid kernel function of the present invention, the curve of the estimated battery SOH value almost coincides with the curve of the true value, and the error is small. However, the curve of the estimated battery SOH value based on the Poly kernel function and the RBF kernel function has a large error.

[0118] Specifically, the error is evaluated using two evaluation indicators: mean absolute percentage error (MAPE) and root mean squared error (RMSE), expressed as follows:

[0119]

[0120]

[0121] Among them, y est and real are the estimated value and true value of the battery SOH, respectively, and n is the number of cycles. When the MAPE and RMSE values ​​are smaller, it means that the estimation error of the battery SOH is smaller, and the performance of the proposed method is better. The specific errors are shown in the table:

[0122]

[0123] As shown in the above table, the battery SOH tested based on the hybrid kernel function is more accurate.

[0124] It can be seen that according to the current curve based on the constant voltage charging stage disclosed in the present invention, the health factors reflecting the battery aging information can be collected, and health factors with stronger applicability and mechanistic rationality can be extracted, and optimization algorithms and data models can be constructed to accurately estimate the health status of the battery and obtain more accurate battery health status results.

[0125] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A battery SOH estimation method based on a constant voltage charging current curve, characterized in that: include: Step S1, during the full cycle aging process of the battery, based on each round of the constant voltage charging current curve, extracting the health factor HI1, the health factor HI2 and the health factor HI3 reflecting the battery aging data and performing normalization processing; Step S2, according to the normalized values ​​of the extracted health factor HI1, the health factor HI2 and the health factor HI3 data, obtaining relevant parameters and training data models based on an optimization algorithm; as well as Step S3, inputting the normalized values ​​of the health factors HI1, HI2 and HI3 into the training data model to calculate the SOH of the battery.

2. The battery SOH estimation method based on constant voltage charging current curve according to claim 1 is characterized in that The step S1 specifically includes: Step S11: pre-processing the capacity data of the battery to obtain the SOH data of the battery: Among them, Q n is the current capacity of the battery, Q o is the rated capacity of the battery; Step S12: extracting the health factors HI1, HI2 and HI3, wherein the health factor HI1 is the time taken for the current to change from a preset current to a target current during the constant voltage charging stage, the health factor HI2 is the charging current turning point information, and the health factor HI3 is the radius of the charging current curve fitting circle; and Step S13: Normalize the health factors HI1, HI2, and HI3: Among them, x' i is the normalized value of the health factors HI1, HI2 and HI3 in the ith cycle, x i are the values ​​of the health factors HI1, HI2 and HI3 in the ith cycle, and min(X) and max(X) are the minimum and maximum values ​​of the health factors HI1, HI2 and HI3 respectively.

3. The battery SOH estimation method based on constant voltage charging current curve according to claim 1, characterized in that: The step S2 specifically includes: Step S21: construct the training data model as a multi-core extreme learning machine model, and introduce the kernel method: f(x')=Hβ Among them, x' is the input sample data, f(x') is the output result, H is the feature mapping matrix of the hidden layer, and β is the weight between the hidden layer and the output layer. Its expression is: Among them, I is the identity matrix, c is the regularization parameter, T is the target vector of the training sample data, and the kernel function matrix model is: H*=HH T Among them, x i ' is a vector consisting of the normalized values ​​of the health factors HI1, HI2 and HI3 of the input feature under the i-th cycle, x' j is the vector composed of the normalized values ​​of the health factors HI1, HI2 and HI3 of the input feature under the j-th cycle, H* is the similarity matrix, and the output result is: Among them, x' is a vector composed of the normalized values ​​of the input features HI1, HI2 and HI3 in a specific cycle, x' n It is a vector composed of the normalized values ​​of the input features HI1, HI2 and HI3 in the nth cycle. Step S22: Optimize using the optimization algorithm to obtain the optimal solution of the relevant parameters required in the multi-core extreme learning machine model.

4. The battery SOH estimation method based on constant voltage charging current curve according to claim 3 is characterized in that: The step S21 also includes combining the Poly kernel function and the RBF kernel function to obtain a hybrid kernel function: K m =ωK RBF +(1-ω)K Poly Where ω is the weight coefficient, K Poly (x',x i ')=(x'·x i '+b) d Among them, x' is a vector composed of the normalized values ​​of the input features HI1, HI2 and HI3 in a specific cycle, x i ' is a vector composed of the normalized values ​​of the input features HI1, HI2 and HI3 in the i-th cycle, b, d are the parameters of the Poly kernel function, and σ is the parameter of the RBF kernel function.

5. The battery SOH estimation method based on constant voltage charging current curve according to claim 3 is characterized in that: The optimization algorithm in step S22 is the Osprey optimization algorithm, which specifically includes: Step S31: Set the number of ospreys T, the maximum number of iterations N, and the parameter range to be optimized; Step S32: Initialize the relevant parameters c, b, d, σ, ω; Step S33: Evaluate the MSE of the fitness SOH training result; Step S34: updating the relevant parameters c, b, d, σ, ω; Step S35: Boundary check; and Step S36: Obtain the optimal solution of the relevant parameters c, b, d, σ, ω.

6. The battery SOH estimation method based on constant voltage charging current curve according to claim 1, characterized in that: The step S3 specifically includes: Step S41: input the normalized values ​​of the health factors HI1, HI2 and HI3; Step S42: Obtaining the optimal solution of the relevant parameters calculated by the optimization algorithm; Step S43: training the data model; Step S44: testing the data model; and Step S45: Output the SOH estimation result of the battery.

7. The battery SOH estimation method based on constant voltage charging current curve according to claim 1, characterized in that: The optimization algorithms include: Osprey optimization algorithm, particle swarm optimization algorithm, grey wolf optimization algorithm and whale optimization algorithm.

8. The battery SOH estimation method based on constant voltage charging current curve according to claim 1, characterized in that: The training data models include: a relevance vector machine model, a support vector machine model, a Gaussian process regression model and a multi-core extreme learning machine model.

Images