Generation and prediction method, device, equipment and medium of battery performance prediction model

By acquiring battery production and performance data, calculating differential curves and feature data, and building a neural network model, the complex and time-consuming problem of existing battery performance prediction methods is solved, and accurate and efficient prediction of battery performance is achieved.

CN118504627BActive Publication Date: 2024-11-05FARASIS TECH (GANZHOU) CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410964230.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-05
Estimated Expiration
2044-07-18

AI Technical Summary

Technical Problem

The existing battery performance prediction methods are complex, time-consuming and energy-consuming, and cannot truly simulate battery performance under different working conditions, and there are problems of accuracy and efficiency.

Method used

By obtaining the battery-formed data and performance data, calculating the differential curve, dividing the capacity interval, extracting feature data, and building a neural network model for training and optimization, a battery performance prediction model is generated.

Benefits of technology

Accurate prediction of battery performance is achieved, prediction accuracy and efficiency are improved, and time and resource consumption of traditional capacity allocation and aging processes are avoided.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118504627B_ABST
    Figure CN118504627B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of battery performance prediction, specifically to a generation and prediction method, device, equipment and medium of a battery performance prediction model, wherein the generation method comprises obtaining battery data; obtaining a differential curve between voltage data and capacity data, and dividing the overall capacity range of multiple batteries into multiple capacity intervals based on multiple differential curves; selecting a specific capacity interval from multiple capacity intervals based on preset rules, and calculating the differential curve value of each specific capacity interval to obtain characteristic data; combining the characteristic data with the corresponding battery performance data to obtain a data set; constructing a neural network model, training the neural network model based on the data set to obtain an initial model, and optimizing the initial model to obtain a battery performance prediction model. The present invention effectively improves the efficiency of predicting battery performance by omitting the battery capacity division and aging process, while reducing the cost of predicting battery performance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of battery performance prediction, and in particular to a method, device, equipment and medium for generating and predicting a battery performance prediction model. Background Art

[0002] Batteries are an important energy storage device that is widely used in various fields. However, as the battery is used for a longer time, its performance gradually decreases, resulting in reduced battery capacity and increased internal resistance. Therefore, how to accurately predict the performance changes of batteries is of great significance to the reliability and life of batteries.

[0003] Battery performance refers to the comprehensive performance of a battery during actual use, and the capacity data of a battery during charge and discharge is one of the important indicators for evaluating changes in battery performance. In the prior art, the evaluation of battery performance usually requires capacity grading and aging processes to confirm the actual performance and life of each battery. However, the traditional capacity grading and aging process has many disadvantages, including complex operating procedures, occupying a large amount of space, consuming a large amount of electricity and time. The capacity grading process usually involves discharging the battery and dividing the battery into different capacity intervals according to its capacity characteristics, which requires a lot of electricity and time, and the operating process is relatively complicated. At the same time, the aging process requires the battery to be left stationary for a long time or to be continuously charged and discharged in cycles to simulate the aging of the battery during actual use. This aging test also requires a lot of time and space, and cannot truly simulate the actual performance of the battery under different working conditions.

[0004] Therefore, the existing technology has certain limitations in predicting battery performance, and there is an urgent need to study a more effective method to predict battery performance. Summary of the invention

[0005] The main purpose of the present invention is to provide a method for generating a battery performance prediction model, aiming to overcome the many shortcomings of existing battery performance prediction methods, including complex operating procedures, occupying a large amount of space, consuming a large amount of electricity and time, and other technical problems.

[0006] In order to solve the above technical problems, the present invention proposes a method for generating a battery performance prediction model, the method comprising:

[0007] A method for generating a battery performance prediction model, comprising:

[0008] Acquiring battery data, wherein the battery data at least includes battery formation data and battery performance data, and the battery formation data includes voltage data and capacity data;

[0009] Acquire a differential curve between the voltage data and the capacity data, and divide the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on a plurality of the differential curves;

[0010] Selecting a specific capacity interval from the plurality of capacity intervals based on a preset rule, and calculating a differential curve value of each specific capacity interval to obtain characteristic data;

[0011] Combining the characteristic data with the corresponding battery performance data to obtain a data set;

[0012] A neural network model is constructed, and the neural network model is trained based on the data set to obtain an initial model, and the initial model is optimized to obtain a battery performance prediction model.

[0013] Furthermore, the step of dividing the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on the plurality of differential curves comprises:

[0014] Plotting a plurality of said differential curves in the same coordinate system;

[0015] The overall capacity range of the plurality of batteries is divided into a plurality of capacity intervals according to the capacity data corresponding to the intersection points of the plurality of differential curves.

[0016] Furthermore, the step of selecting a specific capacity interval from a plurality of capacity intervals based on a preset rule includes:

[0017] Based on the battery formation data, determine whether the battery has a charge and discharge current change or is stationary in each of the capacity intervals;

[0018] If it occurs, the capacity interval in which the battery has the charge and discharge current change or the battery is at rest is excluded to obtain the remaining capacity interval;

[0019] Images of the plurality of differential curves are acquired, and based on the images, a capacity interval with a larger numerical difference is selected from the remaining capacity intervals and set as a specific capacity interval.

[0020] Furthermore, the step of calculating the differential curve value of each specific capacity interval to obtain characteristic data includes:

[0021] The second-order origin moment of the differential curve in each of the specific capacity intervals is calculated to obtain a second-order origin moment data set, wherein the calculation formula of the second-order origin moment is: , is the distribution of the differential curve, is the value of the ith differential curve in a specific capacity interval, is the number of data points within a specific capacity interval;

[0022] The second-order origin moment data set of each specific capacity interval is linearly transformed and normalized to obtain the characteristic data.

[0023] Furthermore, the step of combining the characteristic data with the corresponding battery performance data to obtain a data set includes:

[0024] The characteristic data is used as a characteristic vector and the battery performance data is used as a target vector to obtain a nonlinear regression data set of the battery, wherein the relationship between the characteristic vector and the target vector is expressed as follows: , is the target vector, Represents the nonlinear mapping relationship between the feature vector and the target vector, Number the battery. Indicates the capacity interval, It means that the differential curve is divided into Z capacity intervals. Indicates the number Characteristic data of the second capacity interval of the battery.

[0025] Furthermore, the step of constructing a neural network model includes:

[0026] The neural network model is established based on a multi-layer perceptron algorithm, wherein the neural network model includes an input layer having a nodes, an output layer having one node, and a plurality of hidden layers having a plurality of nodes, wherein a is the number of the specific capacity intervals;

[0027] After the step of building the neural network model, the method further includes:

[0028] Adjusting several nodes of the hidden layer based on a nonlinear activation function;

[0029] Performing a linear connection between the last hidden layer and the output layer;

[0030] The difference between the prediction result of the initial model and the true value is measured based on the relative absolute error function, wherein the expression of the relative absolute error function is:

[0031]

[0032] in is the number of batteries in a certain data set, is the actual performance value of a certain item of the i-th battery, is the performance value of the i-th battery predicted by the initial model.

[0033] Furthermore, before the step of obtaining the battery performance prediction model, the method further includes:

[0034] The initial model is trained and verified based on the training data set and the verification data set to obtain F*H loss values, where F is F groups of training data sets and verification data sets, and H is the number of times training and verification are performed;

[0035] Calculate an average value among the plurality of loss values, and use the average value as a criterion for evaluating the accuracy of the initial model prediction;

[0036] The standard deviation between the multiple loss values ​​is calculated, and the standard deviation is used as a criterion for evaluating the reliability of the initial model prediction.

[0037] The present invention also provides a battery performance prediction method, which performs prediction based on any of the battery performance prediction models described above, and comprises:

[0038] The characteristic data of the battery to be predicted is input into the battery performance prediction model to obtain the actual performance of the battery to be predicted.

[0039] The present invention also provides a battery performance prediction model generation device, comprising:

[0040] A first acquisition module, used for acquiring battery data, wherein the battery data at least includes battery formation data and battery performance data, and the battery formation data includes voltage data and capacity data;

[0041] a second acquisition module, configured to acquire a differential curve between the voltage data and the capacity data, and divide the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on a plurality of the differential curves;

[0042] A selection module, configured to select a specific capacity interval from the plurality of capacity intervals based on a preset rule, and calculate a differential curve value of each specific capacity interval to obtain characteristic data;

[0043] A combining module, used for combining the characteristic data with the corresponding battery performance data to obtain a data set;

[0044] The training module is used to construct a neural network model, train the neural network model based on the data set to obtain an initial model, and optimize the initial model to obtain a battery performance prediction model.

[0045] The present invention further provides a computer device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the computer program.

[0046] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented.

[0047] Beneficial effects:

[0048] This application can more comprehensively describe the state and performance of the battery by using battery formation data and battery performance data, and by dividing the capacity interval based on multiple differential curves and calculating characteristic data. By training and optimizing the neural network model, these characteristic data can be used to accurately predict the performance of the battery and improve the prediction accuracy. When constructing a neural network model, combining the characteristic data with the corresponding battery performance data can make full use of the correlation between the two. By fusing the characteristic data with the battery performance data, the accuracy of the prediction model can be improved, making it more accurately predict the comprehensive performance of the battery, including but not limited to multiple performance indicators such as battery capacity data, providing strong support for the use and maintenance of the battery.

[0049] In summary, the method for generating a battery performance prediction model of the present application combines battery formation data and battery performance data, uses multiple differential curves to divide the capacity interval, and constructs a neural network model for training and optimization, thereby achieving accurate prediction of battery performance without the need for battery capacity division and aging process, effectively saving time and resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a schematic diagram of the steps of a method for generating a battery performance prediction model in one embodiment of the present invention;

[0051] Figure 2 It is a schematic block diagram of the structure of a battery performance prediction model generation device according to an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of the steps of a battery performance prediction method in one embodiment of the present invention;

[0053] Figure 4 A schematic diagram of a specific capacity interval selected according to battery formation data in one embodiment of the present invention;

[0054] Figure 5 A schematic diagram of the difference between characteristic data and capacity data extracted from a capacity interval for multiple batteries in one embodiment of the present invention;

[0055] Figure 6 is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention;

[0056] Figure 7 is a schematic diagram of a neural network structure of an example of the present invention;

[0057] Figure 8 It is characteristic data extracted from the capacity interval of multiple batteries and battery predicted performance data in an example of the present invention.

[0058] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0059] 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.

[0060] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "above", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of features, integers, steps, operations, elements, modules, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, modules, components, and / or groups thereof. It should be understood that when an element is said to be "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any module and all combinations of one or more associated listed items.

[0061] Those skilled in the art will understand that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the field to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0062] refer to Figure 1 The embodiment of the present invention provides a method for generating a battery performance prediction model, comprising the following steps S1-S5:

[0063] S1: Acquire battery data, wherein the battery data at least includes battery formation data and battery performance data, and the battery formation data includes voltage data and capacity data;

[0064] In step S1, the voltage data and capacity data of the battery are collected and processed to obtain voltage data and capacity data. The voltage data refers to the voltage value recorded by the battery during the formation process, and the capacity data refers to the capacity value of the battery under different charge and discharge states. These data can be collected by professional battery testing equipment, such as charge and discharge test instruments. In a preferred embodiment, the battery formation data may include data such as constant current charging current, constant voltage charging current, constant current discharge current, constant voltage discharge current, static voltage and static duration of the battery in multiple different process steps. In addition, it is also necessary to collect battery performance data, which is obtained through laboratory testing or monitoring in actual operation, that is, the battery performance data is data obtained through actual testing. In an optional embodiment, the battery performance data may include initial internal resistance, initial capacity, initial energy, the difference between the initial capacity and the rated capacity of the battery, and the difference between the initial energy and the rated energy; the charge and discharge capacity, charge and discharge energy, and internal resistance of each cycle in the first 200 charge and discharge cycles; the coulomb efficiency, charge and discharge energy efficiency, capacity retention rate, energy retention rate, internal resistance change rate, etc. of the battery in the first 200 charge and discharge cycles. It is worth noting that the battery performance in this embodiment refers to the comprehensive performance exhibited by the battery during actual use, including the capacity data, voltage data, and energy data of the battery during charge and discharge, and the performance indicators such as the battery life and power can be further derived from the capacity data of the battery during charge and discharge. For example, from the capacity data of the battery, the performance changes of the battery in different charge and discharge cycles can be obtained, thereby predicting the battery life, or from the changes in the energy data and voltage data of the battery during charge and discharge, the power changes of the battery can be calculated to further evaluate the performance of the battery in different application scenarios.

[0065] S2: obtaining a differential curve between the voltage data and the capacity data, and dividing the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on a plurality of the differential curves;

[0066] In step S2, by processing the voltage data (V) and capacity data (Q) obtained from the battery formation test, a differential curve (dV / dQ) of the voltage data to the capacity data during the battery formation process can be obtained. The dV / dQ curve describes the change in the voltage of the battery at different capacities and is an important indicator for predicting the battery capacity. The method for calculating the differential curve of the voltage data to the capacity data can be calculated using numerical forward difference, numerical backward difference, numerical central difference and fitted linear regression. After the dV / dQ curve is calculated, it needs to be preprocessed to eliminate noise and outliers, which can be achieved by filtering, smoothing and other methods. Then, by superimposing multiple dV / dQ curves in the same coordinate system, a more accurate capacity prediction result can be obtained. Because of the performance differences of each battery, the superposition of multiple dV / dQ curves can eliminate the influence of the particularity of a single battery and obtain a more general capacity prediction model. And according to the capacity values ​​corresponding to the intersection of the dV / dQ curves, the entire capacity range is divided into multiple intervals. In a preferred embodiment, the dV / dQ curve is analyzed to find the inflection points or extreme points on the curve. These points usually correspond to the critical values ​​of the battery capacity. Mathematical methods or computer algorithms can be used to automatically identify these inflection points or extreme points. According to the inflection points or extreme points found, the overall capacity range is divided into multiple capacity intervals. Each inflection point or extreme point can be used as a demarcation point of an interval, thereby realizing a detailed division of the battery capacity range. In an optional embodiment, the differential curve (dV / dQ) between the voltage data and the capacity data includes: numerical forward difference ( ), numerical backward difference ( ), numerical central difference ( ), fitting linear regression (V = mQ+b, = m), where represents the difference, V represents the voltage data, Q represents the capacity data, and m and b are the coefficients of linear regression.

[0067] S3: selecting a specific capacity interval from the plurality of capacity intervals based on a preset rule, and calculating a differential curve value of each specific capacity interval to obtain characteristic data;

[0068] In step S3, the second-order origin moment is calculated for the differential curve (dV / dQ) in each capacity interval. The second-order origin moment is an indicator used in statistics to describe the shape of data distribution. For curve data, it can reflect the kurtosis or sharpness of the curve. By calculating the second-order origin moment, the curve data can be converted into characteristic data (X). The calculated second-order origin moment is preprocessed, which may include data normalization, outlier processing, feature selection and other steps to ensure the accuracy and stability of the characteristic data. The preprocessed characteristic data is applied to the battery performance prediction model. These characteristic data can be used as the input of the model to help the model better understand the characteristics of the differential curves in different capacity intervals, thereby improving the accuracy of battery performance prediction.

[0069] S4: combining the characteristic data with the corresponding battery performance data to obtain a data set;

[0070] In step S4, the characteristic data obtained in step S3 is combined with the battery performance data obtained in step S1 to form a complete data set. These data sets contain the differential curve characteristics of the battery during the formation process and the performance data obtained from actual tests, providing necessary information for subsequent model training. Subsequently, the data set is divided into a training data set and a validation data set. The training data set is used to train the battery performance prediction model, while the validation data set is used to evaluate the performance of the model. By dividing the data set, it can be ensured that the model will not overfit during the training process, and the results obtained on the validation data set are more representative and generalizable. The above data sets specifically include:

[0071] P(num)= f(µ Inter1 (num), µ Inter2 (num), ..., µ Inter6 (num));

[0072] Where P(num) is the target vector composed of capacity differences (D), f represents the nonlinear mapping relationship between the feature vector and the target vector, num is the battery number, Inter represents the capacity interval, 6 represents that the dV / dQ curve is divided into 6 capacity intervals, and µ Inter2 (num) represents the characteristic data of the second capacity interval of the battery numbered num.

[0073] S5: constructing a neural network model, training the neural network model based on the data set to obtain an initial model, and optimizing the initial model to obtain a battery performance prediction model;

[0074] In step S5, the feature data is merged with the corresponding battery performance data to generate a complete data set, which will be used as the input and verification of the model to ensure the quality and integrity of the data set, including cleaning data, processing missing values, removing outliers, etc. Based on the constructed data set, a neural network model is constructed. A neural network is a mathematical model that imitates the connection between neurons in the human brain. It can identify patterns and associations in the data through learning and training, and then the constructed neural network model is trained using the data set. During the training process, the model will learn and optimize based on the feature data and the corresponding battery performance data to find the best model parameters. This process can include forward propagation (passing the input data through the network to obtain the predicted result) and back propagation (adjusting the model parameters according to the difference between the predicted result and the true value). After obtaining the predicted result in the forward propagation process, the predicted result is compared with the true value, and the value of the loss function is calculated. The loss function measures the difference between the predicted result and the true value, and the goal is to minimize the value of the loss function. After the training is completed, an initial model is obtained, and then further optimized. You can use optimization methods (such as gradient descent algorithm), adjust the model's hyperparameters (such as learning rate, regularization term, etc.) and increase training data. By continuously optimizing the model, you can improve its prediction performance and reduce errors. After each iteration cycle, you can use the validation set to evaluate the performance of the model. By calculating the loss function value and other evaluation indicators of the model on the validation set, you can understand the generalization ability and prediction performance of the model. If the performance of the model on the validation set does not improve or overfitting occurs, you can adjust the model's hyperparameters or take other strategies in time; when the performance of the model on the validation set no longer improves, you can stop training in advance to prevent overfitting. You can set a stop condition, such as the loss function value on the validation set has not significantly improved in several consecutive iterations. After multiple iterations of training and optimization, the final battery performance prediction model is obtained. The model can receive feature data as input and output the corresponding battery performance prediction results. The battery performance prediction model can be used for future battery performance prediction tasks with high accuracy and reliability.

[0075] In a specific embodiment, five battery formation process data are obtained, including specific process steps of the formation process, capacity data (Q), and voltage data (V), the rated capacity of the above batteries and the first discharge capacity of the actual test are obtained, and the capacity difference (D) is calculated, wherein the rated capacity of the battery is the average or median of the amount of electricity that a large number of batteries of this model can provide under standard test conditions, and the capacity difference (D) is the actual first discharge capacity of the battery minus the rated capacity of the battery. Figure 5 The capacity differences of the above five batteries are shown.

[0076] The differential curve (dV / dQ) of the voltage data with respect to the capacity data of the above-mentioned battery in the formation process is calculated by the central difference method, and the above-mentioned differential curve is smoothed using the local weighted regression (LOWESS) method. Then, the dV / dQ curves of the above-mentioned five batteries are superimposed in the same coordinate system, and the entire capacity range is divided into multiple intervals according to the capacity values ​​corresponding to the intersection points of the dV / dQ curves.

[0077] Select the capacity range of 5 batteries and get the following Figure 4 The second-order origin moment of the dV / dQ curve data is calculated in the above 6 capacity intervals ( µ 2) and multiply by 10000 to get Figure 5 The feature data in and combine the feature data into a feature vector (X).

[0078] The characteristic data is combined with the corresponding battery performance data to obtain a data set, specifically including P(num)=f(µ Inter1 (num), µ Inter2 (num), ..., µ Inter6 (num)), and use k-fold cross validation Figure 5 The first round uses the battery number 3840 P(3840) = f(µ Inter1 (3840), µ Inter2 (3840), ..., µ Inter6 (3840)) is the validation set, and the data of the remaining 4 batteries are used as the training set. Similarly, a new battery number is replaced as the validation set in each round, and this process is repeated until all 5 rounds are completed, and 5 sets of validation sets and training sets are obtained.

[0079] Build as Figure 7 The multi-layer perceptron model shown uses the training data set to preliminarily train the neural network model to obtain the initial model. The specific structure of the model is:

[0080] In the input layer, for each battery number num, a 6-dimensional feature vector consisting of battery feature data is received , where each element µ Interi (num) means the battery numbered num is in i Characteristic data of capacity intervals;

[0081]

[0082] Define the output of the first hidden layer as A 1 , which can be expressed as:

[0083] A 1= g( W 1 + B 1 ), W 1 , B 1 , A 1 ; where g() represents the ReLU activation function, N Represents the number of nodes in the current hidden layer, W 1 represents the weight matrix between the input layer and the first hidden layer, B 1 Bias vector for the first hidden layer.

[0084] For l The output of the hidden layer can be expressed as:

[0085] A l = g(A l-1 W l + B l ), W l , B l , A l ; where g() represents the ReLU activation function, M represent l -1 the number of hidden layer nodes, N Represents the number of nodes in the current hidden layer, W l Representative l hidden layers and l -1 weight matrix between hidden layers, B l No. l The bias vector for the hidden layer.

[0086] The final output layer uses a linear connection and can be expressed as:

[0087] P(num) = A l-1 W out + B out , W l , B l , P(num) is the final output result.

[0088] The loss function of the model is the relative absolute error (RAE), and its calculation formula can be expressed as:

[0089]

[0090] Where 4 is the number of batteries in the training set, For the i The actual capacity difference of each battery (D) is The model predicts i The difference in battery capacity (D).

[0091] Using the five groups of data divided by the k-fold cross-validation method, the model was trained and validated to obtain five loss values. The process of training and validating the model was repeated more than 12 times to obtain 5*12 = 60 loss values, and the average value µ and standard deviation σ of the loss values ​​were calculated. The hyperparameters of the model (number of hidden layers, number of nodes in each hidden layer, learning rate, early stopping strategy, etc.) were adjusted to continuously optimize the model, and the model with the smallest average value + standard deviation (µ + σ) was selected as the battery test model. The test results of the eight adjusted models are shown in Table 1.

[0092] The average value µ, standard deviation σ, and average value + standard deviation of 60 loss values ​​calculated on the validation set for the 8 models with different hyperparameters in Table 1 are .

[0093] Table 1

[0094]

[0095] Based on the calculated mean and standard deviation of the loss values ​​of model No. 8 in the validation set, it can be roughly determined that the relative error of the battery performance prediction based on this model is 32.3%, and the probability that the prediction error is within the range of 32.3%±(3*24.9%) is 99.7%. By determining the mean and standard deviation between the loss values ​​of the battery performance prediction model, the distribution of the loss values ​​of the battery performance prediction model can be effectively evaluated. In practical applications, it can help researchers and engineers quickly identify battery performance prediction models with potential problems, thereby providing a basis for further optimizing the model. Figure 5 The data of the five batteries in are all used as the final training data to train the No. 8 model in Table 1 above to obtain the battery performance prediction model.

[0096] In one embodiment, the step of dividing the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on the plurality of differential curves comprises:

[0097] Plotting a plurality of said differential curves in the same coordinate system;

[0098] The overall capacity range of the plurality of batteries is divided into a plurality of capacity intervals according to the capacity data corresponding to the intersection points of the plurality of differential curves.

[0099] In the above embodiment, differential curves of multiple batteries are drawn in the same coordinate system, and the differential curve of each battery is calculated based on its capacity data. The differential curve reflects the rate of change of the battery capacity and can be used to determine the capacity range and interval of the battery. By processing the voltage data V and capacity data Q obtained from the battery formation test, a differential curve (dV / dQ) of the voltage data with respect to the capacity data during the battery formation process can be obtained. The dV / dQ curve describes the change of the battery voltage at different capacities and is an important indicator for predicting the battery capacity. The method for calculating the differential curve of the voltage data with respect to the capacity data can be calculated using numerical forward difference, numerical backward difference, numerical central difference and fitted linear regression.

[0100] In one embodiment, after the step of obtaining a differential curve between the voltage data and the capacity data, the method further includes:

[0101] performing outlier processing on the differential curve based on local outlier detection;

[0102] Filling the missing values ​​in the differential curve based on linear interpolation;

[0103] The differential curve is smoothed based on a smoothing algorithm.

[0104] In the above embodiment, outliers in the differential curve are identified based on an outlier detection algorithm. Outliers may be caused by measurement errors or other external interferences. This can be done by deleting outliers or replacing them with other reasonable values. The specific processing method will depend on the cause of the outliers and the characteristics of the data. Then, linear interpolation is used to fill in the missing values ​​in the differential curve. Missing values ​​may be caused by errors in the data acquisition process or equipment failures. Missing values ​​can be estimated based on the linear relationship between known data points. Finally, the differential curve is smoothed. Smoothing helps to remove noise and unnecessary fluctuations in the curve, making the curve smoother and more readable. Smoothing algorithms including moving average, exponential smoothing or Fibonacci smoothing can be used for smoothing.

[0105] In one embodiment, the step of selecting a specific capacity interval from a plurality of capacity intervals based on a preset rule includes:

[0106] Based on the battery formation data, determine whether the battery has a charge and discharge current change or is stationary in each of the capacity intervals;

[0107] If it occurs, the capacity interval in which the battery has the charge and discharge current change or the battery is at rest is excluded to obtain the remaining capacity interval;

[0108] Images of the plurality of differential curves are acquired, and based on the images, a capacity interval with a larger numerical difference is selected from the remaining capacity intervals and set as a specific capacity interval.

[0109] In the above embodiments, reference Figure 4 , Figure 5 as well as Figure 8 , the battery formation data will be used to determine whether the battery in each capacity interval has a change in charge and discharge current or the battery is stationary. These changes are caused by physical, chemical changes or other factors inside the battery. If the battery has a change in charge and discharge current or the battery is stationary in a certain capacity interval, the capacity interval will be excluded and the remaining capacity interval will be obtained. Multiple differential curve images will be obtained, each battery corresponds to a differential curve, and based on the obtained differential curve images, the capacity interval with a large numerical difference in the remaining capacity interval is selected and set as a specific capacity interval. For each differential curve in the remaining capacity interval, the difference in their numerical values ​​is calculated. These differences can be calculated by various methods, such as calculating the mean square error, correlation coefficient or other statistical methods, and the degree of difference of each differential curve can be obtained, such as Figure 4 As shown, the differences between different differential curves are compared to determine which capacity intervals have the largest differences. The differences can be visually seen by plotting the differential curves. For example, Figure 5 There are 6 voltage intervals shown, among which, Figure 4 The 3840 curve in is the differential curve of the battery numbered 3840. Figure 4 The 4156 curve in is the differential curve of the battery numbered 4156, and so on. Figure 4 In a specific embodiment, the capacity interval with large value difference refers to Figure 4The area with large difference in coordinate data on the y-axis in the , and the area where the coordinate value deviation of the y-axis of a single curve is greater than 1% of the average value of the y-axis coordinate values ​​of multiple curves are capacity intervals with large differences. More specifically, firstly, the average value of the y-axis coordinate values ​​of all differential curves in the remaining capacity interval is calculated. This average value represents the power level of battery charging in this capacity interval. For each differential curve, the deviation of its y-axis coordinate value from the average value is calculated, and a threshold is set. For example, this threshold can be 1% of the average value of the y-axis coordinate values ​​of all differential curves. If the y-axis coordinate value deviation of any differential curve in a capacity interval is greater than this threshold, then this capacity interval is considered to be a capacity interval with large value differences.

[0110] In one embodiment, the step of calculating the differential curve value of each specific capacity interval to obtain characteristic data includes:

[0111] The second-order origin moment of the differential curve in each of the specific capacity intervals is calculated to obtain a second-order origin moment data set, wherein the calculation formula of the second-order origin moment is: , is the distribution of the differential curve, is the value of the ith differential curve in a specific capacity interval, is the number of data points within a specific capacity interval;

[0112] The second-order origin moment data set of each specific capacity interval is linearly transformed and normalized to obtain the characteristic data.

[0113] In the above embodiment, within a specific capacity interval, each differential curve is calculated to obtain its second-order origin moment, which is one of the statistical characteristics of the differential curve and reflects the distribution of the differential curve. Specifically, for the ith differential curve within a specific capacity interval, the calculation formula for its second-order origin moment is: , the second-order origin moment data is formed into a data set, which is a two-dimensional matrix, in which each row represents a differential curve in a specific capacity interval, and each column represents the value of the second-order origin moment. Then, the second-order origin moment data set is linearly transformed and normalized. The purpose of linear transformation is to map the second-order origin moment data set to a new space, so that the model parameters are more stable during the training process. The purpose of normalization is to eliminate the dimensional differences between different features and avoid extreme values ​​covering other data information. Specifically, the second-order origin moment data set can be centered first, that is, each element is subtracted from the mean of the column elements. Then, a linear transformation can be performed. Finally, the transformed data set can be standardized or normalized, for example, it can be scaled to between 0 and 1. After completing the above steps, the feature data is obtained. Feature data refers to the second-order origin moment data set after linear transformation and normalization. These feature data can be further used for classification and clustering analysis to help better understand the performance characteristics of the battery.

[0114] In one embodiment, the step of combining the characteristic data with the corresponding battery performance data to obtain a data set includes:

[0115] The characteristic data is used as a characteristic vector and the battery performance data is used as a target vector to obtain a nonlinear regression data set of the battery, wherein the relationship between the characteristic vector and the target vector is expressed as follows: , is the target vector, Represents the nonlinear mapping relationship between the feature vector and the target vector, Number the battery. Indicates the capacity interval, It means that the differential curve is divided into Z capacity intervals. Indicates the number Characteristic data of the second capacity interval of the battery.

[0116] In the above embodiment, the characteristic data is used as a characteristic vector, and the battery performance data is used as a target vector to obtain a nonlinear regression data set of the battery. This can be achieved by combining the characteristic data and the battery performance data in a certain order. The characteristic vector represents the performance characteristic information of the battery, such as voltage, capacity and other data, and the target vector represents the expected prediction result, that is, the performance prediction of the battery. In the combination process, the corresponding index or label can be used to ensure the correspondence between the characteristic vector and the target vector. It is necessary to establish a nonlinear mapping relationship expression between the characteristic vector and the target vector. A neural network model can be constructed and combined with various nonlinear regression models to achieve this, such as polynomial regression, support vector regression, etc. By selecting a suitable regression model and adjusting relevant parameters, the best nonlinear mapping relationship expression can be obtained to accurately predict the performance of the battery to the maximum extent.

[0117] In another embodiment, the nonlinear regression data set is divided into multiple groups of training data sets and verification data sets based on K-fold cross-validation method or rolling cross-validation method, wherein the training data set is used to train the neural network model, and the verification data set is used to verify the initial model. Specifically, after obtaining the nonlinear regression data set of the battery, further processing and verification of the data set can be achieved by K-fold cross-validation method or rolling cross-validation method. The K-fold cross-validation method divides the data set into K subsets of equal size, wherein K-1 subsets are used for model training, and the remaining subset is used for model verification. K groups of training data sets and verification data sets can be obtained to make full use of the information in the data set and verify the performance of the model. In a specific embodiment, reference Figure 4When the k-fold cross-validation method is used to divide the data set, the first round uses the battery numbered 3840 P(3840) = f(µInter1(3840), µInter2(3840), ..., µInter6(3840)) as the validation set, and the data of the remaining 4 batteries are used as the training set; and so on, a new battery number is replaced as the validation set in each round, and this process is repeated until all 5 rounds are completed, and 5 sets of validation sets and training sets are obtained. Another method is the rolling cross-validation method, which is similar to the division and validation of time series data. In the rolling cross-validation method, the data set is divided into multiple continuous subsets in a certain time sequence, and each subset contains a certain amount of data. Then, each subset is used as a validation data set in turn, and the remaining subsets are used as training data sets, which can simulate the data flow in real situations and verify the model's ability to predict future data. After the data set is divided, the training data set can be used to train the neural network model, and the validation data set can be used to validate the initial model. By continuously adjusting the parameters and structure of the model, the model can be optimized and its performance and accuracy can be improved. At the same time, various evaluation indicators can also be used to measure the performance of the model, such as mean square error, determination coefficient, etc. Through the above steps, the feature data can be combined with the corresponding battery performance data, a nonlinear mapping relationship can be established, and the battery data set can be divided and the model can be verified using the cross-validation method.

[0118] In one embodiment, the step of constructing a neural network model includes:

[0119] The neural network model is established based on a multi-layer perceptron algorithm, wherein the neural network model includes an input layer having a nodes, an output layer having one node, and a plurality of hidden layers having a plurality of nodes, wherein a is the number of the specific capacity intervals;

[0120] After the step of building the neural network model, the method further includes:

[0121] Adjusting several nodes of the hidden layer based on a nonlinear activation function;

[0122] Performing a linear connection between the last hidden layer and the output layer;

[0123] The difference between the prediction result of the initial model and the true value is measured based on the relative absolute error function, wherein the expression of the relative absolute error function is:

[0124]

[0125] in is the number of batteries in a certain data set, is the actual performance value of a certain item of the i-th battery, is the performance value of the i-th battery predicted by the initial model.

[0126] In the above embodiment, based on the multi-layer perceptron algorithm, a neural network model can be constructed, which includes an input layer with a nodes, an output layer of a node, and multiple hidden layers containing several nodes. Wherein, a represents the number of specific capacity intervals. By selecting the appropriate number of nodes and the number of hidden layers, an appropriate neural network structure can be established according to different problems and data sets. After constructing the neural network model, the nonlinear activation function is adjusted for several nodes of the hidden layer. The nonlinear activation function can help the neural network model learn nonlinear relationships and enhance the expression ability of the model. Nonlinear activation functions including Sigmoid function, ReLU function, Tanh function, etc. can be used. By adjusting the parameters and structure of the activation function, the performance of the neural network model can be optimized, and then the last hidden layer is linearly connected to the output layer, and the output of the hidden layer is directly used as the input of the output layer without any nonlinear transformation, which can make the model simpler, reduce the computational complexity, and improve the speed and accuracy of prediction. After building a complete neural network model, the relative absolute error function is needed to measure the difference between the prediction results of the initial model and the true value. The relative absolute error function can reflect the accuracy and stability of the model prediction. In the relative absolute error function, the number of batteries in a certain data set is used as the summation range. For each battery, the difference between its true performance value and a certain performance value predicted by the initial model is calculated and added up. The smaller the final relative absolute error value, the closer the prediction result of the initial model is to the true value.

[0127] In one embodiment, before the step of obtaining the battery performance prediction model, the method further includes:

[0128] The initial model is trained and verified based on the training data set and the verification data set to obtain F*H loss values, where F is F groups of training data sets and verification data sets, and H is the number of times training and verification are performed;

[0129] Calculate an average value among the plurality of loss values, and use the average value as a criterion for evaluating the accuracy of the initial model prediction;

[0130] The standard deviation between the multiple loss values ​​is calculated, and the standard deviation is used as a criterion for evaluating the reliability of the initial model prediction.

[0131] In the above embodiment, the training data set is used to train the parameters of the model, and the validation data set is used to evaluate the performance of the model. The initial model is trained and validated using the training data set and the validation data set. The training and validation process can use methods such as cross-validation to make full use of the data and evaluate the performance of the model. In the training and validation process, F*H loss values ​​are obtained, where F represents the number of groups of training data sets and validation data sets, and H represents the number of times training and validation are performed. For each group of training data sets and validation data sets, we will get a loss value, totaling F*H. The average value of these loss values ​​is calculated, all loss values ​​are added, and the sum is divided by F*H to obtain the average loss value. The average value can be used as a standard for evaluating the accuracy of the initial model prediction, reflecting the average level of the overall performance of the model. In addition to calculating the average value, the standard deviation of these loss values ​​is also calculated. The standard deviation measures the discreteness of the loss value, that is, the dispersion of the loss value relative to its average value. The standard deviation can be used as a standard for evaluating the reliability of the initial model prediction, reflecting the stability and consistency of the model prediction results.

[0132] In a specific embodiment, 5 sets of training sets and validation sets are used to train and validate the model, and 5 loss values ​​are obtained. The steps of training and validating the model are repeated 12 times to obtain 60 loss values, and the average and standard deviation of the loss values ​​are calculated. The number of hidden layers of the model, the number of nodes in each hidden layer, the learning rate, the early stopping strategy, etc. are adjusted, and the model with the smallest sum of the average and standard deviation is selected as the optimized model. The test results of the 8 adjusted models are shown in Table 1.

[0133] According to the calculated mean and standard deviation of the loss values ​​of model 8 in the validation set, it can be roughly determined that the relative error of the battery performance prediction based on this model is 24.9%, and the probability that the prediction error is within the range of 24.9%±(3×57.2%) is 99.7%. By determining the mean and standard deviation between the loss values ​​of the battery performance prediction model, the loss value distribution of the battery performance prediction model can be effectively evaluated. In practical applications, it can help researchers and engineers quickly identify battery performance prediction models with potential problems, thereby providing a basis for further optimizing the model.

[0134] refer to Figure 3 , a battery performance prediction method, the prediction method is based on the above-mentioned battery performance prediction model to predict, comprising:

[0135] S6: Inputting characteristic data of the battery to be predicted into the battery performance prediction model to obtain the actual performance of the battery to be predicted.

[0136] In this embodiment, the characteristic data of the battery to be predicted is input into the established battery performance prediction model. The model will analyze and calculate the input characteristic data to obtain the actual performance prediction result of the battery to be predicted. This result can be performance indicators such as battery capacity, cycle life, and internal resistance. In a preferred embodiment, for example, there is an electric vehicle manufacturing company that wants to use a new battery technology in a new electric vehicle. They have collected a large amount of battery characteristic data and hope to use the model to evaluate the performance of the new battery. First, the characteristic data of the new battery is input into the prediction model. The model will analyze and calculate these characteristic data based on the existing training data and algorithm rules. Finally, the model outputs the predicted performance results of the new battery, such as predicted capacity, cycle life, etc. Based on these prediction results, the electric vehicle manufacturing company can evaluate whether the new battery meets their needs and the performance that may be shown in actual applications.

[0137] In a specific embodiment, the battery performance prediction model obtained in step S7 is used to predict two batteries for which real performance data has not been collected. The characteristic data extracted from the formation process of the two batteries (numbers: 1564, 8156) are input and input into the final model. The model output predicts the capacity difference (D) of the two batteries as follows: Figure 8 shown.

[0138] refer to Figure 2 , a battery performance prediction model generation device, comprising:

[0139] A first acquisition module 100, used to acquire battery data, wherein the battery data at least includes battery formation data and battery performance data, and the battery formation data includes voltage data and capacity data;

[0140] A second acquisition module 200, configured to acquire a differential curve between the voltage data and the capacity data, and divide the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on a plurality of the differential curves;

[0141] A selection module 300, configured to select a specific capacity interval from the plurality of capacity intervals based on a preset rule, and calculate a differential curve value of each specific capacity interval to obtain characteristic data;

[0142] A combining module 400, used to combine the characteristic data with the corresponding battery performance data to obtain a data set;

[0143] The training module 500 is used to construct a neural network model, train the neural network model based on the data set to obtain an initial model, and optimize the initial model to obtain a battery performance prediction model.

[0144] In this embodiment, the present application can more comprehensively describe the state and performance of the battery by using battery formation data and battery performance data, and by dividing the capacity interval based on multiple differential curves and calculating characteristic data. By training and optimizing the neural network model, these characteristic data can be used to accurately predict the performance of the battery and improve the prediction accuracy. When constructing a neural network model, combining the characteristic data with the corresponding battery performance data can make full use of the correlation between the two. By fusing the characteristic data with the battery performance data, the accuracy of the prediction model can be improved, so that it can more accurately predict the performance of the battery.

[0145] Furthermore, the second acquisition module 200 includes:

[0146] A drawing unit, used for drawing a plurality of said differential curves in the same coordinate system;

[0147] The dividing unit is used to divide the overall capacity range of multiple batteries into multiple capacity intervals according to the capacity data corresponding to the intersection points of multiple differential curves.

[0148] Furthermore, the battery performance prediction model generating device further includes:

[0149] A first processing module, configured to perform outlier processing on the differential curve based on local outlier detection;

[0150] A filling module, used for filling the missing values ​​in the differential curve based on a linear interpolation method;

[0151] The second processing module is used to perform smoothing processing on the differential curve based on a smoothing algorithm.

[0152] Furthermore, the selection module 300 includes:

[0153] A first judgment unit, configured to judge whether a charge / discharge current change occurs in the battery or the battery is at rest in each of the capacity intervals based on the battery formation data;

[0154] an exclusion unit, configured to exclude the capacity interval in which the battery has the charge and discharge current change or the battery is at rest, if any, to obtain a remaining capacity interval;

[0155] The first acquisition unit is configured to acquire images of a plurality of the differential curves, and select a capacity interval with a larger value difference from the remaining capacity interval based on the images, and set the capacity interval as a specific capacity interval.

[0156] Furthermore, the selection module 300 further includes:

[0157] The first calculation unit is used to calculate the second-order origin moment of the differential curve in each of the specific capacity intervals to obtain a second-order origin moment data set, wherein the calculation formula of the second-order origin moment is: , is the value of the ith differential curve in a specific capacity interval, is the number of data points within a specific capacity interval;

[0158] The processing unit is used to perform linear transformation and normalization processing on the second-order origin moment data set of each specific capacity interval to obtain the characteristic data.

[0159] Furthermore, the combination module 400 includes:

[0160] A setting unit is used to use the characteristic data as a characteristic vector and the battery performance data as a target vector to obtain a nonlinear regression data set of the battery, wherein the relationship between the characteristic vector and the target vector is expressed as: , is the target vector, Represents the nonlinear mapping relationship between the feature vector and the target vector, Number the battery. Indicates the capacity interval, It means that the differential curve is divided into Z capacity intervals. Indicates the number Characteristic data of the second capacity interval of the battery;

[0161] The battery performance prediction model generating device further includes:

[0162] A splitting module is used to divide the nonlinear regression data set into multiple training data sets and verification data sets based on a cross-validation method, wherein the training data set is used to train the neural network model, and the verification data set is used to verify the initial model.

[0163] Furthermore, the training module 500 includes:

[0164] An establishing unit, used for establishing the neural network model based on a multi-layer perceptron algorithm, wherein the neural network model includes an input layer having a nodes, an output layer having one node, and a plurality of hidden layers having a plurality of nodes, wherein a is the number of the specific capacity intervals;

[0165] Furthermore, the battery performance prediction model generating device further includes:

[0166] An adjustment module, used for adjusting several nodes of the hidden layer based on a nonlinear activation function;

[0167] A linear connection module, used for linearly connecting the last hidden layer to the output layer;

[0168] The third acquisition module is used to measure and obtain the difference between the prediction result of the initial model and the true value based on the relative absolute error function, wherein the expression of the relative absolute error function is:

[0169]

[0170] in is the number of batteries in a certain data set, is the actual performance value of a certain item of the i-th battery, is the performance value of the i-th battery predicted by the initial model.

[0171] Furthermore, the battery performance prediction model generating device further includes:

[0172] A training and verification module is used to train and verify the initial model based on the training data set and the verification data set to obtain F*H loss values, where F is F groups of training data sets and verification data sets, and H is the number of times training and verification are performed;

[0173] A first calculation module, used for calculating an average value between the plurality of loss values, and taking the average value as a criterion for evaluating the accuracy of the prediction of the initial model;

[0174] The second calculation module is used to calculate the standard deviation between the multiple loss values, and use the standard deviation as a criterion for evaluating the reliability of the initial model prediction.

[0175] Furthermore, the present application also discloses a battery performance prediction device, which performs prediction based on any of the battery performance prediction models described above, and the prediction device comprises:

[0176] The performance prediction module is used to input characteristic data of the battery to be predicted into the battery performance prediction model to obtain the actual performance of the battery to be predicted.

[0177] refer to Figure 6 In the embodiment of the present application, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 6As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor designed by the computer 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 memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as a battery performance prediction model. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a battery performance prediction model generation method is implemented.

[0178] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, a method for generating a battery performance prediction model is implemented, comprising the following steps: obtaining battery data, wherein the battery data at least includes battery formation data and battery performance data, and the battery formation data includes voltage data and capacity data; obtaining a differential curve between the voltage data and the capacity data, and dividing the overall capacity range of multiple batteries into multiple capacity intervals based on multiple differential curves; selecting a specific capacity interval from the multiple capacity intervals based on preset rules, and calculating the differential curve value of each specific capacity interval to obtain feature data; combining the feature data with the corresponding battery performance data to obtain a data set; constructing a neural network model, training the neural network model based on the data set to obtain an initial model, and optimizing the initial model to obtain a battery performance prediction model.

[0179] 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 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 provided in this application and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0180] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for generating a battery performance prediction model, characterized in that: include: Acquiring battery data, wherein the battery data at least includes battery formation data and battery performance data, and the battery formation data includes voltage data and capacity data; Acquire a differential curve between the voltage data and the capacity data, and divide the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on a plurality of the differential curves; Selecting a specific capacity interval from the plurality of capacity intervals based on a preset rule, and calculating a differential curve value of each specific capacity interval to obtain characteristic data; Combining the characteristic data with the corresponding battery performance data to obtain a data set; Constructing a neural network model, training the neural network model based on the data set to obtain an initial model, and optimizing the initial model to obtain a battery performance prediction model; The step of calculating the differential curve value of each specific capacity interval to obtain characteristic data includes: The second-order origin moment of the differential curve in each of the specific capacity intervals is calculated to obtain a second-order origin moment data set, wherein the calculation formula of the second-order origin moment is: , is the distribution of the differential curve, is the value of the ith differential curve in a specific capacity interval, is the number of data points within a specific capacity interval; The second-order origin moment data set of each specific capacity interval is linearly transformed and normalized to obtain the characteristic data.

2. The method for generating a battery performance prediction model according to claim 1, characterized in that: The step of dividing the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on the plurality of differential curves comprises: Plotting a plurality of said differential curves in the same coordinate system; The overall capacity range of the plurality of batteries is divided into a plurality of capacity intervals according to the capacity data corresponding to the intersection points of the plurality of differential curves.

3. The method for generating a battery performance prediction model according to claim 1, characterized in that: The step of selecting a specific capacity interval from a plurality of capacity intervals based on a preset rule includes: Based on the battery formation data, determine whether the battery has a charge and discharge current change or is stationary in each of the capacity intervals; If it occurs, the capacity interval in which the battery has the charge and discharge current change or the battery is at rest is excluded to obtain the remaining capacity interval; Images of the plurality of differential curves are acquired, and based on the images, a capacity interval with a larger numerical difference is selected from the remaining capacity intervals and set as a specific capacity interval.

4. The method for generating a battery performance prediction model according to claim 1, characterized in that: The step of combining the characteristic data with the corresponding battery performance data to obtain a data set includes: The characteristic data is used as a characteristic vector and the battery performance data is used as a target vector to obtain a nonlinear regression data set of the battery, wherein the relationship between the characteristic vector and the target vector is expressed as follows: , is the target vector, Represents the nonlinear mapping relationship between the feature vector and the target vector, Number the battery. Indicates the capacity interval, It means that the differential curve is divided into Z capacity intervals. Indicates the number Characteristic data of the second capacity interval of the battery.

5. The method for generating a battery performance prediction model according to claim 1, characterized in that: The step of constructing the neural network model includes: The neural network model is established based on a multi-layer perceptron algorithm, wherein the neural network model includes an input layer having a nodes, an output layer having one node, and a plurality of hidden layers having a plurality of nodes, wherein a is the number of the specific capacity intervals; After the step of building the neural network model, the method further includes: Adjusting several nodes of the hidden layer based on a nonlinear activation function; Performing a linear connection between the last hidden layer and the output layer; The difference between the prediction result of the initial model and the true value is measured based on the relative absolute error function, wherein the expression of the relative absolute error function is: in is the number of batteries in a certain data set, is the actual performance value of a certain item of the i-th battery, is the performance value of the i-th battery predicted by the initial model.

6. The method for generating a battery performance prediction model according to claim 4, characterized in that: Before the step of obtaining the battery performance prediction model, the method further includes: The initial model is trained and verified based on the training data set and the verification data set to obtain F*H loss values, where F is F groups of training data sets and verification data sets, and H is the number of times training and verification are performed; Calculate an average value among the plurality of loss values, and use the average value as a criterion for evaluating the accuracy of the initial model prediction; The standard deviation between the multiple loss values ​​is calculated, and the standard deviation is used as a criterion for evaluating the reliability of the initial model prediction.

7. A battery performance prediction method, the prediction method is based on the battery performance prediction model according to any one of claims 1 to 6, characterized in that: include: The characteristic data of the battery to be predicted is input into the battery performance prediction model to obtain the actual performance of the battery to be predicted.

8. A battery performance prediction model generation device, used to implement the method according to any one of claims 1 to 6, characterized in that: include: A first acquisition module, used for acquiring battery data, wherein the battery data at least includes battery formation data and battery performance data, and the battery formation data includes voltage data and capacity data; a second acquisition module, configured to acquire a differential curve between the voltage data and the capacity data, and divide the overall capacity range of the plurality of batteries into a plurality of capacity intervals based on a plurality of the differential curves; A selection module, configured to select a specific capacity interval from the plurality of capacity intervals based on a preset rule, and calculate a differential curve value of each specific capacity interval to obtain characteristic data; A combining module, used for combining the characteristic data with the corresponding battery performance data to obtain a data set; The training module is used to construct a neural network model, train the neural network model based on the data set to obtain an initial model, and optimize the initial model to obtain a battery performance prediction model.

9. A computer device, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.