Prediction method and device of discharge capacity, model training method and device, equipment and medium

Abstract

The invention provides a discharge capacity prediction method and device, a model training method and device, equipment and a medium, and the model training method comprises the steps: obtaining a data sample set which comprises a feature data sample of a negative electrode material sample and a discharge capacity data sample, the discharge capacity data sample is obtained by performing an electrochemical test on the negative electrode material sample; the characteristic data samples in the data sample set are used as input, the discharge capacity data samples in the data sample set are used as output, at least one basic model is trained to obtain a target prediction model, the discharge capacity of the negative electrode material can be directly predicted through the target prediction model, and compared with traditional electrochemical testing, the operation is simple, and the efficiency is high. And the consumption of manpower, material resources and time cost is greatly saved.

Description

Technical Field

[0001] The present disclosure relates to the field of batteries, and in particular to a discharge capacity prediction method, model training method, device, equipment, and medium. Background Art

[0002] The high energy density, high volume density, and relatively high operating voltage of lithium-ion batteries have become the cornerstone of the success of electric vehicles and a sustainable energy economy.

[0003] In lithium-ion batteries, the discharge capacity of the negative electrode material significantly impacts the overall battery capacity. Therefore, to determine the discharge capacity of negative electrode materials, manufacturers currently typically use traditional electrochemical testing. This involves assembling the negative electrode material into a battery and then testing it with a specialized battery performance testing system. However, this testing method is cumbersome and inconvenient to manufacture. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the defect that the discharge capacity testing method of the negative electrode material in the prior art is too cumbersome, and to provide a discharge capacity prediction method, model training method, device, equipment and medium.

[0005] The present invention solves the above technical problems through the following technical solutions:

[0006] In a first aspect, the present disclosure provides a model training method, the method comprising:

[0007] Acquire a data sample set, the data sample set including a characteristic data sample and a discharge capacity data sample of the negative electrode material sample, the discharge capacity data sample being obtained by performing an electrochemical test on the negative electrode material sample;

[0008] Taking the characteristic data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output, at least one basic model is trained to obtain a target prediction model.

[0009] In an optional embodiment, at least one basic model is trained using characteristic data samples in a data sample set as input and discharge capacity data samples in the data sample set as output to obtain a target prediction model, including:

[0010] Preprocessing the feature data samples in the data sample set, where the preprocessing includes at least one of data dimensionality reduction, data cleaning, and data normalization;

[0011] Taking the preprocessed feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output, at least one basic model is trained to obtain a target prediction model.

[0012] In an optional embodiment, the negative electrode material sample includes characteristic data samples of at least two dimensions;

[0013] Preprocess the feature data samples in the data sample set, including:

[0014] For the negative electrode material sample, determining the correlation coefficient between the characteristic data samples of each two dimensions;

[0015] When the absolute value of the correlation coefficient is greater than a first preset threshold, the feature data samples that meet the preset filtering conditions in the feature data samples of the two dimensions corresponding to the correlation coefficient are filtered out.

[0016] In an optional embodiment, the correlation coefficient includes a Pearson correlation coefficient;

[0017] And / or, the characteristic data sample includes D of the artificial graphite material min 、D 10 、D 50 、D 90 、D 99 、D max , tap density, specific surface area, apparent density, degree of graphitization, powder OI value, magnetic foreign matter, pole piece compaction density, volatile matter of raw coke, ash content of raw coke, moisture content of raw coke, sulfur content of raw coke, grindability index of raw coke, bonding index of raw coke and particle sphericity. At least two dimensions.

[0018] In an optional embodiment, at least one basic model is trained using characteristic data samples in a data sample set as input and discharge capacity data samples in the data sample set as output to obtain a target prediction model, including:

[0019] Training at least one basic model using feature data samples in the data sample set as input and discharge capacity data samples in the data sample set as output;

[0020] A base model that meets the model evaluation condition in at least one base model is determined as a target prediction model.

[0021] In an optional embodiment, the data sample set is divided into a training subset and a test subset;

[0022] Taking the feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output, training at least one basic model includes:

[0023] Training at least one basic model using the feature data samples in the training subset as input and the discharge capacity data samples in the training subset as output;

[0024] Determining a base model that meets a model evaluation condition in at least one base model as a target prediction model includes:

[0025] At least one basic model is tested using a test subset, and a basic model that meets the model evaluation conditions is determined as a target prediction model.

[0026] In an optional embodiment, the model evaluation conditions include:

[0027] a determination coefficient of the basic model being greater than or equal to a second preset threshold, the determination coefficient being calculated based on the discharge capacity data samples in the test subset and an output value of the basic model;

[0028] And / or, a mean square error of the basic model is less than or equal to a third preset threshold, where the mean square error is calculated based on the discharge capacity data samples in the test subset and an output value of the basic model.

[0029] In a second aspect, the present disclosure provides a method for predicting discharge capacity, the method comprising:

[0030] Obtaining characteristic data of the negative electrode material to be tested;

[0031] The characteristic data is input into a target prediction model to obtain the discharge capacity of the negative electrode material to be tested, and the target prediction model is trained according to any model training method in the first aspect.

[0032] In a third aspect, the present disclosure provides a model training device, characterized in that the device includes:

[0033] A sample acquisition module is used to acquire a data sample set, the data sample set including a characteristic data sample and a discharge capacity data sample of a negative electrode material sample, the discharge capacity data sample being obtained by performing an electrochemical test on the negative electrode material sample;

[0034] The model training module is used to train at least one basic model using the feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output to obtain a target prediction model.

[0035] In a fourth aspect, the present disclosure provides a discharge capacity prediction device, the device comprising:

[0036] A data acquisition module is used to obtain characteristic data of the negative electrode material to be tested;

[0037] The capacity determination module is used to input the characteristic data into the target prediction model to obtain the discharge capacity of the negative electrode material to be tested, and the target prediction model is trained according to the model training device of any one of the third aspects.

[0038] In a fifth aspect, the present disclosure provides an electronic device, comprising a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the model training method as described in any one of the first aspects, or the discharge capacity prediction method as described in the second aspect.

[0039] In a sixth aspect, the present disclosure provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the model training method as described in any one of the first aspects, or the discharge capacity prediction method as described in the second aspect.

[0040] The positive progress of this disclosure is:

[0041] This disclosure trains at least one basic model using characteristic data samples and discharge capacity data samples as input and output, respectively, to generate a target prediction model. The target prediction model can directly predict the discharge capacity of the negative electrode material based solely on the characteristic data of the negative electrode material. Compared to traditional electrochemical testing, directly predicting the discharge capacity of the negative electrode material using the target prediction model is simple to operate and significantly saves manpower, material resources, and time costs.

[0042] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of a model training method provided by an exemplary embodiment of the present disclosure;

[0044] Figure 2 A first flow chart of step S102 of a model training method provided by an exemplary embodiment of the present disclosure;

[0045] Figure 3 A flowchart of step S1021 of a model training method provided by an exemplary embodiment of the present disclosure;

[0046] Figure 4 A heat diagram showing the correlation coefficients of feature data samples of each two dimensions provided by an exemplary embodiment of the present disclosure;

[0047] Figure 5 A second flow chart of step S102 of a model training method provided by an exemplary embodiment of the present disclosure;

[0048] Figure 6A schematic diagram showing a comparison of the coefficients of determination of three basic models on a training subset and a test subset provided by an exemplary embodiment of the present disclosure;

[0049] Figure 7 A schematic diagram showing a comparison of mean square errors of three basic models on training subsets and test subsets provided in an exemplary embodiment of the present disclosure;

[0050] Figure 8 A schematic diagram of accuracy verification of an extreme gradient boosting model provided by an exemplary embodiment of the present disclosure;

[0051] Figure 9 A schematic flow chart of a method for predicting discharge capacity provided by an exemplary embodiment of the present disclosure;

[0052] Figure 10 A schematic diagram of a module of a model training device provided by an exemplary embodiment of the present disclosure;

[0053] Figure 11 A schematic diagram of a module of a discharge capacity prediction device provided by an exemplary embodiment of the present disclosure;

[0054] Figure 12 A schematic structural diagram of an electronic device provided by an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present disclosure in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0056] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0057] The following describes a model training method provided by an embodiment of the present disclosure. Figure 1It is a flowchart of a model training method provided by an embodiment of the present disclosure. This specification provides method operation steps such as the embodiment or flowchart, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiment is only one way of executing the steps among many steps, and does not represent the only execution order. When the actual system or server product is executed, it can be executed in sequence or in parallel (for example, in a parallel processor or multi-threaded processing environment) according to the method shown in the embodiment or the accompanying drawings. Specifically, Figure 1 As shown, the method may include:

[0058] S101: Obtain a data sample set.

[0059] The data sample set is used to train the basic model so that the trained model can predict the discharge capacity of the negative electrode material sample based on the characteristic data of the negative electrode material. The data sample set includes characteristic data samples and discharge capacity data samples of the negative electrode material sample.

[0060] Specifically, the characteristic data sample is a material characteristic that can affect the discharge capacity sample, and the characteristic data sample may generally include at least one dimension.

[0061] Taking the negative electrode material sample as artificial graphite as an example, the characteristic data sample may include at least one of the following dimensions: D min (minimum particle size), D 10 (particle size corresponding to 10% of cumulative distribution), D 50 (median particle size), D 90 (particle size corresponding to 90% of the cumulative distribution), D 99 (particle size corresponding to 99% of cumulative distribution), D max (maximum particle size), tap density (TD), specific surface area (BET), apparent density (AD), degree of graphitization (GD), powder OI value (OI), magnetic foreign matter (MF), pole piece compaction density (CD), volatile matter of raw coke (VM), ash content of raw coke (AC), moisture content of raw coke (WC), sulfur content of raw coke (SC), grindability index (GI) of raw coke, adhesion index (CI) of raw coke and particle sphericity, etc., but not limited to these, and can be selected according to actual conditions.

[0062] Specifically, the discharge capacity data sample refers to the amount of electricity that can be released by the negative electrode material sample during the discharge process, and the discharge capacity data sample is obtained by performing an electrochemical test on the negative electrode material sample.

[0063] Taking artificial graphite as an example, to conduct electrochemical testing on the negative electrode material sample, the negative electrode material sample first needs to be assembled into a battery, and then the assembled battery is tested for charge and discharge capacity on a battery testing system to obtain a discharge capacity data sample. The operation of assembling the negative electrode material into a battery specifically includes the following steps:

[0064] Step 1, homogenization, that is, mixing the negative electrode material, conductive agent, binder and solvent evenly to form a slurry; Step 2, coating, that is, evenly coating the homogenized slurry on the current collector (such as aluminum foil or copper foil); Step 3, rolling, that is, compacting the coated electrode through a roller press to improve the density and consistency of the electrode; Step 4, drying, that is, removing the solvent in the electrode to solidify the electrode material; Step 5, cutting, cutting the dried electrode into the required shape and size; Step 6, weighing, that is, measuring the weight of the electrode sheet to ensure that the active material content of each electrode is consistent; Step 7, button battery assembly, that is, assembling the electrode sheet, diaphragm and electrolyte into a button battery (button battery).

[0065] S102 : Using the feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output, train at least one basic model to obtain a target prediction model.

[0066] Regarding the basic model, a selection can be made based on the basic model's ability to process high-dimensional data and nonlinear relationships. In this embodiment, the following three models are preferably used as the basic models of the disclosed embodiment: random forest model (RFR), gradient boosting model (GBR), and extreme gradient boosting model (XGBoost), but are not limited to these. The specific selection can be made based on the actual application scenario. The above is only an example.

[0067] The random forest model is an ensemble learning method that combines multiple decision trees into a more powerful model. In a random forest regression model, each decision tree is fitted based on a specific subsample of the data and randomly selected features. This means that random forests are able to capture the diverse characteristics of the training data and avoid overfitting.

[0068] The gradient boosting model draws on the strengths of models such as decision trees, random forests, adaptive boosting (Adaboost), and gradient descent, continuously improving their effectiveness. The basic idea behind the gradient boosting model is a residual learning approach, where new models are continuously added to reduce the model's residuals. The core idea is to add new models to fit the residuals of the current model and assign appropriate weights to each new model, ensuring that it plays a more significant role in the overall learning process.

[0069] The extreme gradient boosting model is an ensemble learning algorithm based on decision trees. Its core concept is to use a special cost function in the first step of learning to distinguish more valuable data. This valuable data is then processed more carefully during tree construction. In extreme gradient boosting, the cost function consists of two parts: the cost of model complexity and the cost of prediction errors. The cost of model complexity uses regularization to prevent overfitting. The cost of prediction errors uses an exponential loss function to emphasize the gap between the predicted value and the true value.

[0070] The basic model is selected based on the performance characteristics of the model, aiming to cover algorithm levels from simple to complex, so as to determine the performance of different basic models on the data sample set, so as to facilitate the screening of the target prediction model that better meets user needs.

[0071] In this embodiment, the characteristic data samples and the discharge capacity data samples are used as the input and output of the basic model respectively to train at least one basic model to obtain a target prediction model. The target prediction model can directly predict the discharge capacity of the negative electrode material according to the characteristic data of the negative electrode material. Compared with the traditional electrochemical test, it greatly saves the consumption of manpower, material resources and time costs; at the same time, it avoids the use of chemical reagents that may be used in traditional electrochemical tests, such as N-methylpyrrolidone, lithium hexafluorophosphate, etc., which greatly improves the safety of the experiment, reduces the generation of waste materials and waste liquids, and is environmentally friendly.

[0072] In an alternative embodiment, see Figure 2 , step S102 specifically includes:

[0073] S1021. Preprocess the feature data samples in the data sample set.

[0074] Preprocessing includes at least one of data dimensionality reduction, data cleaning, and data normalization. The following are some specific examples of data preprocessing operations that can be taken:

[0075] Data dimensionality reduction may include but is not limited to: principal component analysis, which reduces feature dimensions through linear changes while retaining as many main features as possible; linear discriminant change, which reduces feature dimensions by maximizing inter-category variance and minimizing intra-category variance; and correlation coefficient dimensionality reduction, which analyzes the correlation between each two feature indicators through the correlation coefficient to eliminate redundant features.

[0076] Data cleaning may include but is not limited to: missing value processing, that is, filling (median, mean, mode) or deleting feature data samples with missing values; outlier processing, that is, detecting and processing abnormal feature data samples through interquartile range or Z-SCORE method.

[0077] Data normalization may include but is not limited to: Min-Max normalization, which scales the data to [0, 1] or [-1, 1]; and Z-SCORE normalization, which converts the data into a standard normal distribution with a mean of 0 and a standard deviation of 1.

[0078] The preprocessing operations are not limited to the above examples and can be selected according to actual conditions.

[0079] S1022: Using the preprocessed feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output, train at least one basic model to obtain a target prediction model.

[0080] In this embodiment, in order to improve the training accuracy during the model training process and reduce the interference of input samples on the model training, the acquired feature data samples can be preprocessed before the basic model is trained.

[0081] Furthermore, the negative electrode material sample includes characteristic data samples in at least two dimensions.

[0082] During the model training process, when the dimension of the feature data samples used as model input increases, the performance of the basic model may improve until the optimal number of elements is reached. At this time, if the dimension of the feature data samples is further increased, the distribution of the feature data samples in the high-dimensional space will become very sparse, and the basic model will find it difficult to find meaningful patterns in the high-dimensional space, which may easily lead to overfitting in the training of the basic model. In this embodiment, the correlation coefficient is used to perform data dimensionality reduction processing, see Figure 3 , step S1021 includes:

[0083] S10211. For each negative electrode material sample, determine the correlation coefficient between the characteristic data samples of each two dimensions.

[0084] The correlation coefficient may be a Pearson correlation coefficient, but is not limited thereto.

[0085] Assume that every two feature data samples are and , the formula for the Pearson correlation coefficient is as follows:

[0086]

[0087] in, is the Pearson correlation coefficient, and The kth sample is respectively and The value of and The feature data samples are and The mean of , N is the number of samples.

[0088] S10212: When the absolute value of the correlation coefficient is greater than a first preset threshold, filtering out the feature data samples that meet the preset filtering conditions in the feature data samples of the two dimensions corresponding to the correlation coefficient.

[0089] The filtering preset conditions include selecting the less important feature data samples in the two dimensions. There are many ways to consider the importance of feature data samples, such as by staff based on the laws of battery negative electrode materials, through electrochemical testing, or by calculating and determining the feature data samples using a random forest algorithm. In some application scenarios, the importance of two feature data samples can also be determined by training models based on the feature data samples separately and comparing the performance of the trained models.

[0090] In addition, the setting of the first preset threshold can be selected according to specific circumstances. In this embodiment, it is preferably 0.8, but not limited to this.

[0091] Generally speaking, if the absolute value of the correlation coefficient is less than 0.5, it means that the correlation between the feature data samples of the corresponding two dimensions is low; if the absolute value of the correlation coefficient is between 0.5-0.8, it means that the correlation between the feature data samples of the corresponding two dimensions is acceptable; if the absolute value of the correlation coefficient is less than 0.8, it means that the correlation between the feature data samples of the corresponding two dimensions is high.

[0092] For feature data samples with high correlation and meeting the preset filtering conditions, it means that they are of little help in improving the accuracy of model training, but are more likely to cause the risk of model overfitting. Filtering this type of feature data samples can avoid overfitting during model training and shorten the calculation time of model training.

[0093] The following uses artificial graphite as an example of anode material, but the specific implementation is not limited to this example. The characteristic data sample includes D min 、D 10 、D 50 、D 90 、D 99 、D max , tap density, specific surface area, apparent density, degree of graphitization, powder OI value, magnetic foreign matter, pole piece compaction density, volatile matter of raw coke, ash content of raw coke, moisture content of raw coke, sulfur content of raw coke, grindability index of raw coke, bonding index of raw coke and particle sphericity. At least two dimensions.

[0094] Calculate the correlation coefficient between the feature data samples of each two dimensions in the above feature data samples, and screen out the feature data samples of the two dimensions corresponding to the correlation coefficient greater than 0.8. In this embodiment, the correlation coefficient between the particle sphericity and the tap density is greater than 0.8. At this time, according to the importance of the two feature data samples of particle sphericity and tap density, the feature data sample with lower importance in the two feature data samples is filtered out. In this embodiment, the basic model is trained with particle sphericity and tap density as input samples, and the performance of the basic model is judged. It is found that the performance of the basic model trained by tap density is better. Therefore, the feature data sample of particle sphericity is filtered out, and the feature data sample of tap density is retained.

[0095] Finally, the feature data samples after data dimensionality reduction include: D min 、D 10 、D 50 、D 90 、D 99 、D max , tap density, specific surface area, apparent density, degree of graphitization, powder OI value, magnetic foreign matter, pole piece compaction density, volatile matter of raw coke, ash content of raw coke, moisture content of raw coke, sulfur content of raw coke, grindability index of raw coke, and bonding index of raw coke. Figure 4 , re-evaluate the correlation of the feature data samples after data dimensionality reduction, Figure 4 The part circled in red represents the maximum correlation coefficient of 0.44, which is less than the first preset threshold of 0.8. This indicates that the correlations between the above feature data samples are relatively low, meeting the model training requirements.

[0096] In an optional embodiment, in order to evaluate the training process of the basic model and select the most suitable target prediction model, see Figure 5 , step S102 further includes:

[0097] S1023. Train at least one basic model using the feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output.

[0098] S1024: Determine the basic model that meets the model evaluation conditions among the at least one trained basic model as the target prediction model.

[0099] The model evaluation criteria are used to verify and evaluate the predictive performance of each base model. The specific settings can be selected based on the actual application scenario. Generally speaking, meeting the model evaluation criteria indicates that the fitting accuracy is the best among at least one base model, and the fitting accuracy can meet the requirements of discharge capacity prediction.

[0100] In an optional embodiment, the data sample set is divided into a training subset and a test subset. The training subset is used to train the basic model, and the test subset is used to test the trained basic model. Specifically:

[0101] Step S1023 further includes: training at least one basic model using the feature data samples in the training subset as input and the discharge capacity data samples in the training subset as output.

[0102] Step S1024 also includes: testing at least one basic model through a test subset, and determining a basic model that meets the model evaluation conditions as a target prediction model.

[0103] Specifically, the model evaluation condition includes: the determination coefficient of the basic model is greater than or equal to a second preset threshold, and the determination coefficient is calculated based on the discharge capacity data samples in the test subset and the output value of the basic model:

[0104]

[0105] in, represents the discharge capacity data sample, represents the output value of the basic model, The coefficient of determination score ranges from 0 to 1. The closer it is to 1, the higher the accuracy of the model prediction. The second preset threshold can be set according to the actual situation.

[0106] In addition, in this embodiment, the algorithm for calculating the coefficient of determination is embedded in the base model. This means that the base model can directly output the coefficient of determination based on the input data sample set. Furthermore, because the algorithm for calculating the coefficient of determination is embedded in the base model, the corresponding coefficient of determination is actually output during the training of the base model using the training subset.

[0107] See also Figure 6 Taking the random forest model (RFR), gradient boosting model (GBR), and extreme gradient boosting model (XGBoost) as the basic models as examples, the mean square error comparison diagrams of the above three models on the training subset (Train) and the test subset (Test) are shown respectively, and the Y axis shows the value of the respective determination coefficient (R 2 Score), by comparison, it can be found that the extreme gradient boosting model has the largest determination coefficient both on the training subset and the test subset, which meets the prediction requirements. Therefore, in this embodiment, the trained extreme gradient boosting model can be determined as the target prediction model.

[0108] Specifically, the model evaluation conditions include: the mean square error of the basic model is less than or equal to a third preset threshold, and the mean square error is calculated based on the discharge capacity samples in the test subset and the output value of the basic model:

[0109]

[0110] in, represents the discharge capacity data sample, represents the output value of the basic model, The closer the mean square error is to 0, the higher the accuracy of the model prediction. The third preset threshold can be set according to the actual situation.

[0111] In addition, in this embodiment, the algorithm for calculating the mean square error is embedded in the base model. This means that the base model can directly output the mean square error based on the input data sample set. Furthermore, since the algorithm for calculating the mean square error is embedded in the base model, the corresponding mean square error is actually output during the training of the base model using the training subset.

[0112] See also Figure 7 Taking the random forest model (RFR in the figure), the gradient boosting model (GBR in the figure), and the extreme gradient boosting model (XGBoost in the figure) as basic models as examples, the mean square error comparison diagrams of the above three models on the training subset (Train in the figure) and the test subset (Test in the figure) are respectively shown. The Y-axis shows the respective mean square error values ​​(MSE in the figure). Through comparison, it can be found that the extreme gradient boosting model has the smallest mean square error both in the training subset and the test subset, which meets the prediction requirements. Therefore, in this embodiment, the trained extreme gradient boosting model can be determined as the target prediction model.

[0113] In addition, in order to further verify the accuracy of the extreme gradient lifting model, the present disclosure is verified by the discharge capacity data sample of the negative electrode material, and the verification results are as follows: Figure 8As shown. The X-axis represents the discharge capacity data samples obtained from electrochemical testing (i.e., Actual testing in the figure), the Y-axis represents the output value of the extreme gradient boosting model (i.e., MLpredicions in the figure), the blue sample points represent the data obtained in the training subset, and the green sample points represent the data obtained in the test subset. The black dotted line in the figure is X=Y, which represents whether there is a good consistency between the output value of the extreme gradient boosting model and the discharge capacity data samples obtained from the electrochemical test. The farther the data point is from the black dotted line, the greater the error. Conversely, the closer the data point is to the black dotted line, the smaller the error. It is not difficult to find from the figure that whether in the training subset or the test subset, the distribution of the output value of the extreme gradient boosting model is basically consistent with the distribution of the discharge capacity data samples of the electrochemical test, which proves that the XGboost model has good application performance in predicting battery capacity.

[0114] An exemplary embodiment of the present disclosure further provides a method for predicting discharge capacity, see Figure 9 , methods include:

[0115] S901. Obtain characteristic data of the negative electrode material to be tested.

[0116] The characteristic data are usually material parameters that can affect the discharge capacity of the negative electrode material to a certain extent, and the characteristic data can usually include characteristic data of at least one dimension.

[0117] Taking artificial graphite as the negative electrode material as an example, the characteristic data may include the following dimensions: D min 、D 10 、D 50 、D 90 、D 99 、D max , tap density (TD), specific surface area (BET), apparent density (AD), degree of graphitization (GD), powder OI value (OI), magnetic foreign matter (MF), pole piece compaction density (CD), volatile matter of raw coke (VM), ash content of raw coke (AC), moisture content of raw coke (WC), sulfur content of raw coke (SC), grindability index (GI) of raw coke, adhesion index (CI) of raw coke and particle sphericity, etc.

[0118] S902 , inputting the characteristic data into a target prediction model to obtain the discharge capacity of the negative electrode material to be tested.

[0119] The discharge capacity data of the negative electrode material refers to the amount of electricity that the negative electrode material can release during the discharge process, and the target prediction model is trained according to the model training method in the above embodiment.

[0120] An exemplary embodiment of the present disclosure provides a model training device. Figure 10, the device comprises:

[0121] The sample acquisition module 1001 is used to acquire a data sample set, which includes characteristic data samples and discharge capacity data samples of negative electrode material samples. The discharge capacity data samples are obtained by performing electrochemical tests on the negative electrode material samples.

[0122] The model training module 1002 is configured to train at least one basic model using the feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output to obtain a target prediction model.

[0123] In an optional embodiment, the model training module 1002 includes:

[0124] The preprocessing unit is used to preprocess the feature data samples in the data sample set, and the preprocessing includes at least one of data dimensionality reduction, data cleaning and data normalization.

[0125] The first training unit is used to train at least one basic model using the preprocessed feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output to obtain a target prediction model.

[0126] In an optional embodiment, the negative electrode material sample includes characteristic data samples in at least two dimensions.

[0127] The preprocessing unit is further used to determine the correlation coefficient between the characteristic data samples of each two dimensions for the negative electrode material sample; when the absolute value of the correlation coefficient is greater than a first preset threshold, the characteristic data samples that meet the preset filtering conditions in the characteristic data samples of the two dimensions corresponding to the correlation coefficient are filtered out.

[0128] In an optional embodiment, the correlation coefficient includes a Pearson correlation coefficient.

[0129] In an optional embodiment, the characteristic data sample includes D min、 D 10 、D 50 、D 90 、D 99 、D max , tap density, specific surface area, apparent density, degree of graphitization, powder OI value, magnetic foreign matter, pole piece compaction density, volatile matter of raw coke, ash content of raw coke, moisture content of raw coke, sulfur content of raw coke, grindability index of raw coke, bonding index of raw coke and particle sphericity. At least two dimensions.

[0130] In an optional embodiment, the model training module 1002 includes:

[0131] a first training unit, configured to train at least one basic model using feature data samples in the data sample set as input and discharge capacity data samples in the data sample set as output;

[0132] An evaluation unit is used to determine a basic model that meets the model evaluation conditions in at least one basic model as a target prediction model.

[0133] In an optional embodiment, the data sample set is divided into a training subset and a test subset;

[0134] The first training unit is further configured to train at least one basic model using the feature data samples in the training subset as input and the discharge capacity data samples in the training subset as output;

[0135] The evaluation unit is further used to test at least one basic model through a test subset, and determine the basic model that meets the model evaluation conditions as the target prediction model.

[0136] In an optional embodiment, the model evaluation conditions include:

[0137] a determination coefficient of the basic model being greater than or equal to a second preset threshold, the determination coefficient being calculated based on the discharge capacity data samples in the test subset and an output value of the basic model;

[0138] In an optional embodiment, the model evaluation conditions include:

[0139] The mean square error of the basic model is less than or equal to a third preset threshold, and the mean square error is calculated based on the discharge capacity data samples in the test subset and the output value of the basic model.

[0140] An exemplary embodiment of the present disclosure provides a device for predicting discharge capacity. Figure 11 , the device comprises:

[0141] The data acquisition module 1101 is used to acquire characteristic data of the negative electrode material to be tested.

[0142] The capacity determination module 1102 is used to input the characteristic data into a target prediction model to obtain the discharge capacity of the negative electrode material to be tested. The target prediction model is trained according to the model training method in the above embodiment.

[0143] Figure 12 This is a structural diagram of an electronic device provided in an exemplary embodiment of the present disclosure. Figure 12 The electronic device 1200 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0144] like Figure 12As shown, the electronic device 1200 may be a general-purpose computing device, such as a server device. Components of the electronic device 1200 may include, but are not limited to, the at least one processor 1201, the at least one memory 1202, and a bus 1203 connecting different system components (including the memory 1202 and the processor 1201).

[0145] The bus 1203 includes a data bus, an address bus, and a control bus.

[0146] The memory 1202 may include a volatile memory, such as a random access memory (RAM) 12021 and / or a cache memory 12022 , and may further include a read-only memory (ROM) 12023 .

[0147] The memory 1202 may also include a program tool 1202120 (or utility) having a set (at least one) of program modules 12024, such program modules 12024 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0148] The processor 1201 executes various functional applications and data processing by running the computer program stored in the memory 1202, such as the method provided in any of the above embodiments.

[0149] Electronic device 1200 can also communicate with one or more external devices 1204. This communication can occur via input / output (I / O) interface 1205. Furthermore, model-generated electronic device 1200 can also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via network adapter 1206. As shown, network adapter 1206 communicates with other modules of model-generated electronic device 1200 via bus 1203. It should be understood that, although not shown, other hardware and / or software modules can be used in conjunction with model-generated electronic device 1200, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID (RAID) systems, tape drives, and data backup storage systems.

[0150] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0151] An embodiment of the present disclosure further provides a computer-readable storage medium having a computer program stored thereon, which implements the method provided in any of the above embodiments when the program is executed by a processor.

[0152] The readable storage medium may include, but is not limited to, a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0153] In a possible implementation manner, the embodiments of the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute a method for implementing any of the above embodiments.

[0154] The program code for executing the present disclosure may be written in any combination of one or more programming languages, and may be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0155] While specific embodiments of the present disclosure have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of protection of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, and such changes and modifications are intended to fall within the scope of protection of the present disclosure.

Claims

1. A model training method, characterized in that: The method comprises: Acquire a data sample set, the data sample set including a characteristic data sample and a discharge capacity data sample of a negative electrode material sample, the discharge capacity data sample being obtained by performing an electrochemical test on the negative electrode material sample; Taking the characteristic data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output, at least one basic model is trained to obtain a target prediction model.

2. The model training method according to claim 1, wherein: The method of training at least one basic model using the characteristic data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output to obtain a target prediction model includes: Preprocessing the feature data samples in the data sample set, wherein the preprocessing includes at least one of data dimensionality reduction, data cleaning, and data normalization; Taking the preprocessed feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output, at least one basic model is trained to obtain a target prediction model.

3. The model training method according to claim 2, wherein: The negative electrode material sample includes characteristic data samples of at least two dimensions; The preprocessing of the characteristic data samples in the data sample set includes: For the negative electrode material sample, determining the correlation coefficient between the characteristic data samples of each two dimensions; When the absolute value of the correlation coefficient is greater than a first preset threshold, the feature data samples that meet the preset filtering conditions in the feature data samples of the two dimensions corresponding to the correlation coefficient are filtered out.

4. The model training method according to claim 3, wherein: The correlation coefficient includes the Pearson correlation coefficient; And / or, the characteristic data sample includes D of the artificial graphite material min 、D 10 、D 50 、D 90 、D 99 、D max , tap density, specific surface area, apparent density, degree of graphitization, powder OI value, magnetic foreign matter, pole piece compaction density, volatile matter of raw coke, ash content of raw coke, moisture content of raw coke, sulfur content of raw coke, grindability index of raw coke, bonding index of raw coke and particle sphericity. At least two dimensions.

5. The model training method according to claim 1, wherein: The method of training at least one basic model using the characteristic data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output to obtain a target prediction model includes: Training at least one basic model using the feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output; A base model that meets the model evaluation condition among the at least one base model is determined as the target prediction model.

6. The model training method according to claim 5, wherein: The data sample set is divided into a training subset and a test subset; The training of at least one basic model using the feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output includes: Training at least one basic model using the feature data samples in the training subset as input and the discharge capacity data samples in the training subset as output; The step of determining a base model that satisfies a model evaluation condition among the at least one base model as a target prediction model includes: The at least one basic model is tested using the test subset, and a basic model that meets the model evaluation conditions is determined as a target prediction model.

7. The model training method according to claim 6, wherein: The model evaluation conditions include: The determination coefficient of the basic model is greater than or equal to a second preset threshold, the determination coefficient being calculated based on the discharge capacity data samples in the test subset and the output value of the basic model; And / or, a mean square error of the basic model is less than or equal to a third preset threshold, and the mean square error is calculated based on the discharge capacity data samples in the test subset and the output value of the basic model.

8. A method for predicting discharge capacity, characterized in that: The method comprises: Obtaining characteristic data of the negative electrode material to be tested; The characteristic data is input into a target prediction model to obtain the discharge capacity of the negative electrode material to be tested, wherein the target prediction model is trained according to the model training method according to any one of claims 1 to 7.

9. A model training device, characterized in that: The device comprises: a sample acquisition module, configured to acquire a data sample set, wherein the data sample set includes a characteristic data sample and a discharge capacity data sample of a negative electrode material sample, wherein the discharge capacity data sample is obtained by performing an electrochemical test on the negative electrode material sample; The model training module is used to train at least one basic model using the feature data samples in the data sample set as input and the discharge capacity data samples in the data sample set as output to obtain a target prediction model.

10. A discharge capacity prediction device, characterized in that: The device comprises: A data acquisition module is used to obtain characteristic data of the negative electrode material to be tested; A capacity determination module is used to input the characteristic data into a target prediction model to obtain the discharge capacity of the negative electrode material to be tested, wherein the target prediction model is trained according to the model training device as described in claim 9.

11. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the model training method according to any one of claims 1 to 7, or the discharge capacity prediction method according to claim 8.

12. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the model training method according to any one of claims 1 to 7, or the discharge capacity prediction method according to claim 8.

Images