Method and equipment for predicting cycle performance of battery and storage medium

Abstract

The invention discloses a method and device for predicting the cycle performance of a battery and a storage medium. The method for predicting the cycle performance of the battery comprises the steps that performance characteristics of the battery to be tested are obtained from early-stage used data of the battery to be tested; and performing prediction processing on the performance characteristics based on a pre-trained cycle performance prediction model to obtain cycle performance indexes of the to-be-tested battery. By adopting the method for predicting the cycle performance of the battery, the influence of the performance of the battery on the cycle performance of the battery can be considered, and the accuracy of predicting the cycle performance of the battery is improved.

Description

Technical Field

[0001] The present application belongs to the field of battery technology, and in particular, relates to a method, device and storage medium for predicting battery cycle performance. Background Art

[0002] With the advancement of social technology, batteries are needed in many fields, such as various power tools, electric vehicles, drones, etc. In terms of battery optimization design, manufacturing and application, battery cycle performance is an important evaluation indicator.

[0003] In order to predict the battery cycle performance, the Arrhenius formula is often used to derive the battery life at different temperatures. However, the Arrhenius formula focuses on the impact of a single external environmental factor on the battery life, resulting in a low accuracy in predicting the battery cycle performance. Summary of the invention

[0004] The embodiments of the present application provide a method, device and storage medium for predicting battery cycle performance, which can improve the prediction accuracy of battery cycle performance.

[0005] In a first aspect, an embodiment of the present application provides a method for predicting battery cycle performance, the method comprising:

[0006] Obtaining performance characteristics of the battery under test from previous usage data of the battery under test;

[0007] The performance characteristics are predicted based on the pre-trained cycle performance prediction model to obtain the cycle performance index of the battery to be tested.

[0008] Optionally, the performance characteristics of the battery to be tested are obtained from the previous usage data of the battery to be tested, including:

[0009] Obtaining the previous usage data of the battery to be tested, the previous usage data including the charge and discharge voltage data, charge and discharge current data, battery capacity data, temperature data and pressure data at each historical cycle number;

[0010] Perform data cleaning on the previously used data to obtain processed previously used data;

[0011] The performance characteristics of the battery to be tested are determined based on the charge and discharge voltage data, charge and discharge current data, and battery capacity data at each historical cycle number in the processed previous usage data.

[0012] Optionally, the estimated results are predicted by the result output layer to obtain the target battery capacity under the target cycle number, including:

[0013] The estimated results are predicted through the result output layer to obtain the predicted battery capacity under the target cycle number;

[0014] Count the average temperature of temperature data under each historical cycle number;

[0015] The predicted battery capacity is adjusted according to the temperature average value to obtain the target battery capacity.

[0016] Optionally, the classification results of each feature are aggregated by an estimator to obtain an estimation result, including:

[0017] Determine the number weight of the historical number of cycles corresponding to each sub-feature classifier through an estimator;

[0018] The feature classification results and frequency weights corresponding to each sub-feature classifier are weighted by the estimator to obtain the weighted classification results corresponding to each sub-feature classifier;

[0019] The weighted classification results are aggregated through the estimator to obtain the estimated result.

[0020] Optionally, the method further includes:

[0021] Obtaining a sample battery training set, the sample battery training set includes a plurality of cycle test feature sets of sample batteries, and the cycle test feature set of each sample battery includes its own performance characteristics of the sample battery tested in each single test cycle;

[0022] Get the regression model to be trained;

[0023] The regression model is trained based on the sample battery training set until the preset conditions are met, and a trained cycle performance prediction model is obtained.

[0024] Optionally, obtain a regression model to be trained, including:

[0025] Based on the network search algorithm, the regression model is traversed to obtain multiple candidate hyperparameter combinations;

[0026] A cross-validation algorithm is used to divide the sample battery data set into a sample battery training set and a sample battery validation set, and the performance of the regression model under each candidate hyperparameter combination is verified to obtain the performance verification result corresponding to each candidate hyperparameter;

[0027] According to the performance verification results, a target hyperparameter combination is determined from multiple candidate hyperparameter combinations;

[0028] Construct a regression model to be trained based on the target hyperparameter combination.

[0029] Optionally, obtaining the performance characteristics of the regression model to be trained includes battery capacity, cycle number, charge and discharge voltage, and charge and discharge current;

[0030] The regression model is trained based on the sample battery training set until the preset conditions are met, and a trained cycle performance prediction model is obtained, including:

[0031] The number of cycles, charge and discharge voltage, and charge and discharge current in each single test cycle in the sample battery training set are used as the input of the regression model, and the battery capacity in each single test cycle is used as the output of the regression model. The regression model is trained until the preset conditions are met to obtain a cycle performance prediction model.

[0032] Optionally, during the training of the regression model, the method further includes:

[0033] Using an evaluation index to evaluate the model performance of the cycle performance prediction model, the evaluation index includes at least one of a mean square error, a root mean square error, and a determination coefficient; or,

[0034] By displaying a scatter plot of the predicted values ​​and true values ​​of the cycle performance prediction model, the model performance of the cycle performance prediction model can be evaluated based on the scatter plot.

[0035] Optionally, before obtaining the sample battery training set, the method further includes:

[0036] Respectively obtain cycle test data of each sample battery within a preset period, wherein each sample battery has a set of single test data in each single test cycle;

[0037] Performing data cleaning processing on the cycle test data of each sample battery to obtain processed cycle test data;

[0038] The performance characteristics of the sample battery are determined from the processed cycle test data to construct a sample battery feature set, and the sample battery training set is selected from the sample battery feature set.

[0039] Optionally, the cycle performance prediction model includes a feature sampling layer, a feature classifier, an estimator and a result output layer connected in sequence, the feature classifier includes a plurality of parallel sub-feature classifiers; the performance characteristics include the historical number of cycles, the historical charge and discharge voltage and the historical charge and discharge current, and the cycle performance index is measured by the battery capacity;

[0040] The performance characteristics are predicted based on the pre-trained cycle performance prediction model to obtain the cycle performance indicators of the battery to be tested, including:

[0041] Based on the feature sampling layer, historical charge and discharge currents and historical charge and discharge voltages at multiple different historical cycle numbers are selected from the performance features as inputs of the feature classifier;

[0042] Each sub-feature classifier performs feature classification processing on a historical charge and discharge current and a historical charge and discharge voltage under a historical cycle number inputted by each sub-feature classifier, and obtains a feature classification result of each sub-feature classifier;

[0043] The classification results of each feature are summarized and processed by the estimator to obtain the estimation result;

[0044] The estimated results are predicted through the result output layer to obtain the target battery capacity under the target number of cycles.

[0045] In a second aspect, an embodiment of the present application provides a device for predicting battery cycle performance, comprising:

[0046] A feature acquisition module, used to acquire the performance features of the battery to be tested from the previous usage data of the battery to be tested;

[0047] The indicator prediction module is used to predict the performance characteristics based on the pre-trained cycle performance prediction model to obtain the cycle performance indicators of the battery to be tested.

[0048] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0049] Processor; and

[0050] The memory stores a computer program, and when the computer program is executed by the processor, the processor executes any one of the above methods.

[0051] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement any of the methods described above.

[0052] The embodiment of the present application takes into account the performance characteristics related to the performance of the battery to be tested, and predicts the performance characteristics through a pre-trained cycle performance prediction model to obtain the cycle performance characteristics of the battery to be tested. Compared with the prior art that only considers the impact of a single external environmental factor on the battery cycle performance, the present application reduces the error in the cycle performance prediction of the battery to be tested by adopting characteristics related to the battery's own performance and excluding characteristics unrelated to the battery's own performance. In addition, the embodiment of the present application also predicts the cycle performance of the battery to be tested by training a cycle performance prediction model, which can effectively process multiple characteristics related to the battery's own performance to improve the accuracy and reliability of the battery cycle performance prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The technical solution and beneficial effects of the present application will be made apparent by describing in detail the specific implementation methods of the present application in conjunction with the accompanying drawings.

[0054] Figure 1 A schematic diagram of a process for predicting battery cycle performance provided in an embodiment of the present application;

[0055] Figure 2 A schematic diagram of the structure of the cycle performance prediction model provided in the embodiment of the present application;

[0056] Figure 3 A schematic diagram of a training cycle performance prediction model in the method provided in an embodiment of the present application;

[0057] Figure 4 A schematic diagram of the structure of a device for predicting battery cycle performance provided in an embodiment of the present application;

[0058] Figure 5 Another schematic diagram of the structure of the device for predicting battery cycle performance provided in an embodiment of the present application;

[0059] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] Please refer to the drawings, where the same component symbols represent the same components. The principle of the present application is illustrated by implementing it in an appropriate computing environment. The following description is based on the illustrated specific embodiments of the present application, which should not be considered as limiting other specific embodiments of the present application that are not described in detail herein.

[0061] In order to improve the accuracy of battery cycle performance prediction, the present application embodiment introduces a machine learning model and predicts the battery cycle performance index by training the machine learning model. To elaborate on the scheme provided by the present application embodiment, the following introduces the method for predicting battery cycle performance provided by the present application embodiment in two aspects: model training and model application.

[0062] For model applications, see Figure 1 , Figure 1 1 is a flow chart of a method for predicting battery cycle performance provided in an embodiment of the present application. The method for predicting battery cycle performance can be applied to a device for predicting battery cycle performance provided in an embodiment of the present application, and the device can also be integrated in an electronic device, which can be a mobile phone, a computer, a tablet, etc.

[0063] The process of the method for predicting battery cycle performance provided in the embodiment of the present application may include:

[0064] In 101 , the performance characteristics of the battery to be tested are obtained from the previous usage data of the battery to be tested.

[0065] The battery for which the battery cycle performance prediction is required is called a test battery, and the test battery may be a lithium-ion battery. After the test battery is used for a period of time, corresponding usage data will be generated, and the usage data is called the previous usage data.

[0066] In some embodiments, obtaining the performance characteristics of the battery to be tested from the previous usage data of the battery to be tested includes:

[0067] Obtaining the previous usage data of the battery to be tested, the previous usage data including the charge and discharge voltage data, charge and discharge current data, battery capacity data, temperature data and pressure data at each historical cycle number;

[0068] Perform data cleaning on the previously used data to obtain processed previously used data;

[0069] The performance characteristics of the battery to be tested are determined based on the charge and discharge voltage data, charge and discharge current data, and battery capacity data at each historical cycle number in the processed previous usage data.

[0070] Exemplarily, the previous used data of the battery to be tested includes data related to its own performance and data unrelated to its own performance. Among them, the data related to the performance of the battery to be tested include battery capacity data, charge and discharge voltage data, charge and discharge current data, charge and discharge power data, historical cycle times (also called charge and discharge usage times), internal resistance data, etc. The data unrelated to the performance of the battery to be tested include temperature data, pressure data, etc.

[0071] After obtaining the previous usage data of the battery to be tested, data related to the performance of the battery to be tested can be filtered out from the previous usage data, such as the number of historical cycles, charge and discharge voltage data, charge and discharge current data, and battery capacity data.

[0072] Thereafter, the screened data may be normalized or standardized to obtain the performance characteristics of the battery to be tested, so that the data related to the performance of the battery to be tested itself are at the same numerical level, so as to serve as the input of the subsequent model.

[0073] In addition, after obtaining the previous usage data of the battery to be tested, the previous usage data can be cleaned to obtain the processed previous usage data. Then, data related to the performance of the battery to be tested can be filtered out from the previous usage data. Finally, the filtered data can be normalized or standardized to obtain the performance characteristics of the battery to be tested.

[0074] Among them, data cleaning processing of previously used data includes partially removing outliers or missing values ​​in previously used data, or correcting outliers and filling missing values.

[0075] In the embodiments of the present application, by obtaining performance characteristics related to the performance of the battery to be tested from the previous usage data of the battery to be tested, on the one hand, the cycle performance of the battery to be tested is predicted by integrating multiple performance characteristics, thereby improving the reliability and stability of the cycle performance prediction of the battery to be tested; on the other hand, the performance characteristics related to the performance of the battery to be tested are used to predict the cycle performance of the battery to be tested, thereby filtering out the influence of irrelevant factors on the prediction results, thereby improving the accuracy of the cycle performance prediction of the battery to be tested.

[0076] In 102, the performance characteristics are predicted based on the pre-trained cycle performance prediction model to obtain the cycle performance index of the battery to be tested.

[0077] In an embodiment of the present application, a machine learning model is first constructed, and then the machine learning model is trained to predict the cycle performance indicators of the battery to be tested.

[0078] Among them, the constructed machine learning model can be a regression model, which is trained using a sample battery training set, and the trained machine learning model is called a cycle performance prediction model.

[0079] The embodiment of the present application can predict the unknown cycle performance index of the battery to be tested based on the known performance characteristics of the battery to be tested by training the regression model, thereby improving the efficiency of predicting the cycle performance of the battery to be tested. It can also comprehensively consider the impact of various complex factors on the cycle performance of the battery, has a higher data processing capability, and can effectively process the correlation between the performance characteristics of the battery to be tested compared to the empirical model and electrochemical model used in the relevant technology, and is generally applicable to the prediction of the cycle performance of most batteries.

[0080] The above empirical model and electrochemical model are explained here. The empirical model is a battery attenuation model constructed through a large amount of battery experimental data. It depends on specific condition types and battery types, and requires high-quality processing of a large amount of battery experimental data. Therefore, it has defects such as poor versatility and difficulty in processing experimental data. The electrochemical model is a mathematical model constructed based on the electrochemical principles and material properties of the battery. It is used to simulate the chemical reactions and physical processes inside the battery to predict the cycle performance of the battery. This type of mathematical model has the defects of high complexity, large amount of calculation, and a large number of electrochemical parameters that are difficult to obtain in practical applications.

[0081] When using the cycle performance prediction model, the performance characteristics are input into the cycle performance prediction model for prediction processing, and the prediction result of the cycle performance of the battery to be tested is output, and the prediction result is represented by a cycle performance index. Among them, the cycle performance index refers to a parameter used to represent the battery cycle performance, such as the number of cycles, battery capacity, etc.

[0082] The embodiment of the present application trains a cycle performance prediction model to predict the cycle performance index of the battery to be tested according to its performance characteristics. The cycle performance prediction model can be applied to most batteries and has high data processing efficiency.

[0083] Here we introduce the cycle performance prediction model provided in the embodiment of this application. Figure 2 , Figure 2 The schematic diagram of the structure of the cycle performance prediction model provided in the embodiment of the present application. The cycle prediction model provided in the embodiment of the present application includes a feature sampling layer, a feature classifier, an estimator and a result output layer connected in sequence, and the feature classifier includes a plurality of parallel sub-feature classifiers.

[0084] Among them, the feature sampling layer is used to extract multiple sub-features from the performance features and randomly input them into multiple sub-feature classifiers, each of which has one sub-feature input. Each sub-feature classifier is used to perform feature classification processing on the sub-features it inputs to obtain feature classification results. The estimator is used to summarize the feature classification results of multiple sub-feature classifiers to obtain estimation results. The result output layer is used to make predictions based on the estimation results.

[0085] Exemplarily, the performance characteristics include historical cycle numbers, historical charge and discharge voltages, and historical charge and discharge currents, and the cycle performance index is measured by the battery capacity.

[0086] The performance characteristics are predicted based on the pre-trained cycle performance prediction model to obtain the cycle performance indicators of the battery to be tested, including:

[0087] Based on the feature sampling layer, historical charge and discharge currents and historical charge and discharge voltages at multiple different historical cycle numbers are selected from the performance features as inputs of the feature classifier;

[0088] Each sub-feature classifier performs feature classification processing on a historical charge and discharge current and a historical charge and discharge voltage under a historical cycle number inputted by each sub-feature classifier, and obtains a feature classification result of each sub-feature classifier;

[0089] The classification results of each feature are summarized and processed by the estimator to obtain the estimation result;

[0090] The estimated results are predicted through the result output layer to obtain the target battery capacity under the target number of cycles.

[0091] Among them, the performance characteristics of the battery to be tested are input into the feature sampling layer, and the feature sampling layer is used to divide the characteristics in the performance characteristics. For example, the historical charge and discharge current in the performance characteristics is represented by In, the historical charge and discharge voltage is represented by Vn, and the historical number of cycles is represented by N. According to the test sequence of the battery to be tested, the performance characteristics can be represented as: (N=1, I1, V1) (N=2, I2, V2)... (N, In, Vn).

[0092] A set of performance features under a test sequence can be randomly input into a sub-feature classifier in the feature classifier through the feature sampling layer. For example, (N=2, I2, V2) is input into the sub-feature classifier F1, and (N=5, I5, V5) is input into the sub-feature classifier F10.

[0093] Then, each sub-feature classifier predicts the corresponding battery capacity as the feature classification result based on its input. The battery capacity is represented by Cn, and n represents the number of historical cycles. The estimator establishes the corresponding relationship between N-In-Vn-Cn based on each predicted battery capacity. The result output layer is used to obtain the battery capacity under the target number of cycles based on the estimation result of the estimator and the target number of cycles. Among them, the value of the target number of cycles is greater than the maximum value of the historical number of cycles.

[0094] In addition, through long-term observation of battery test results, in this embodiment, the performance characteristics of the battery for the first 300 cycles can be selected as the input of the cycle performance prediction model to predict the battery capacity at 600 cycles.

[0095] For example, the structure of the cycle performance prediction model provided in the embodiment of the present application may include a random forest machine learning model. In the process of predicting the battery cycle performance through the model, multiple performance characteristics related to the performance of the battery to be tested can be comprehensively considered, and the law of change of performance characteristics with the number of cycles can be captured, thereby improving the accuracy of the battery cycle performance prediction.

[0096] In some embodiments, the estimation result is predicted by the result output layer to obtain the target battery capacity under the target cycle number, including:

[0097] The estimated results are predicted through the result output layer to obtain the predicted battery capacity under the target cycle number;

[0098] Count the average temperature of temperature data under each historical cycle number;

[0099] The predicted battery capacity is adjusted according to the temperature average value to obtain the target battery capacity.

[0100] The battery to be tested has corresponding temperature data at different historical cycle times, and the temperature average value is obtained by averaging each temperature data, so as to adjust the predicted battery capacity according to the temperature average value. It can be understood that before obtaining the temperature average value, the maximum value or minimum value in each temperature data can be screened out to make the obtained temperature average value closer to the long-term use temperature of the battery to be tested.

[0101] When adjusting the predicted battery capacity according to the temperature average value, the absolute value of the difference between the temperature average value and the preset temperature can be determined first, so as to determine the adjustment range of the predicted battery capacity according to the absolute value of the difference. For example, if the temperature average value is greater than the preset temperature, the predicted battery capacity is reduced to obtain the target battery capacity. The reduction range of the predicted battery capacity is positively correlated with the absolute value of the difference. That is, the larger the absolute value of the difference, the greater the reduction range of the predicted battery capacity, and the target battery capacity is much smaller than the predicted battery capacity; the smaller the absolute value of the difference, the smaller the reduction range of the predicted battery capacity, and the target battery capacity is slightly smaller than the predicted battery capacity.

[0102] In this embodiment, the preset temperature may be the temperature at which the cycle performance of the battery to be tested is optimal during use. The temperature may be obtained through experimental measurement or manual setting, or may be room temperature. The specific setting may be based on actual needs.

[0103] As another embodiment, temperature levels with different temperature average values ​​may be pre-divided, and different temperature levels have corresponding temperature coefficients. By multiplying the temperature coefficient by the predicted battery capacity, the target battery capacity may be obtained. The higher the temperature average value, the lower the temperature level, and the smaller the temperature coefficient, which is less than or equal to 1.

[0104] In this embodiment, the estimation result is first predicted through the result output layer to obtain the predicted battery capacity under the preliminary predicted target cycle number, and then the influence of environmental external factors on the battery to be tested is taken into consideration, and the predicted battery capacity is adjusted in combination with the temperature data of the battery to be tested, so as to accurately evaluate the target battery capacity of the battery to be tested in the actual use environment, so that the obtained target battery capacity value is more accurate, which is convenient for the subsequent accurate evaluation of the cycle performance of the battery to be tested.

[0105] In some embodiments, the estimator aggregates the feature classification results to obtain an estimation result, including:

[0106] Determine the number weight of the historical number of cycles corresponding to each sub-feature classifier through an estimator;

[0107] The feature classification results and frequency weights corresponding to each sub-feature classifier are weighted by the estimator to obtain the weighted classification results corresponding to each sub-feature classifier;

[0108] The weighted classification results are aggregated through the estimator to obtain the estimated result.

[0109] Among them, the number of times weights can be pre-set for different historical cycle times, and the number of times weights are proportional to the numerical value of the historical cycle times. The larger the historical cycle times, the larger the number of times weights, and the smaller the historical cycle times, the smaller the number of times weights. After the performance characteristics under different historical cycle times are input into the corresponding sub-feature classifier for processing, the weighted classification results output by each sub-feature classifier can be weighted by the estimator, and then the weighted classification results after weighted processing are summarized to obtain the estimated results. For example, the number of historical cycles corresponding to sub-feature classifier 1 is the 5th, and the output weighted classification result is a1, and the number of times weight corresponding to the 5th historical cycle is f5, and a1*f5 can be used as the weighted classification result corresponding to the 5th historical cycle (sub-feature classifier 1).

[0110] In addition, the gradient difference of each weight of each number can also be determined according to the value of the target number of cycles. For example, the larger the target number of cycles, the smaller the gradient difference of each weight of each number, and the smaller the target number of cycles, the larger the gradient difference of each weight of each number.

[0111] This embodiment takes into account the target battery capacity under the target cycle number that needs to be predicted later, and is most affected by the feature classification results under the historical cycle number closest to the target cycle number. Therefore, by setting corresponding number weights for different historical cycle numbers and assigning weights to different feature classification results through the estimator, the accuracy of the estimation result can be greatly improved.

[0112] For model training, see Figure 3 , Figure 3 A schematic diagram of a training cycle performance prediction model in the method provided in an embodiment of the present application includes the following process:

[0113] In 201, a sample battery training set is obtained, the sample battery training set includes a plurality of cycle test feature sets of sample batteries, and the cycle test feature set of each sample battery includes its own performance characteristics of the sample battery tested in each single test cycle.

[0114] For example, the test data of some sample batteries are extracted from the test database in advance to create a sample battery training set. There are multiple sample batteries, and each sample battery has multiple sets of cycle test feature sets. Each set of cycle test feature sets includes test cycle number N, charge and discharge voltage Vn, charge and discharge current In, battery capacity Cn and other features related to the performance of the sample battery itself. Each single test cycle corresponds to a charge and discharge voltage, charge and discharge power, and battery capacity.

[0115] In some embodiments, the method of preparing a sample battery training set includes:

[0116] 2011. Obtaining cycle test data of each sample battery within a preset period, respectively, wherein each sample battery has a set of single test data in each single test cycle.

[0117] For example, the preset cycle represents the duration from the 0th cycle test to the 600th cycle test. If the single test data of each time is represented by the cycle test, then the 600 cycle tests of each sample battery will generate 600 sets of single test data, and each single test data may include charge and discharge current, charge and discharge voltage, battery capacity, temperature, pressure, etc.

[0118] 2012. Perform data cleaning on the cycle test data of each sample battery to obtain processed cycle test data.

[0119] Among them, data cleaning processing includes processing outliers, missing values, etc. By removing the part of the N groups of single test data of each sample battery that includes outliers or missing values, or correcting the outliers and filling the missing values, the cycle test data after processing of each sample battery is obtained. In this way, the data quality of the cycle test data is improved, so that the cycle test data can achieve better results when applied.

[0120] 2013. From the processed cycle test data, the performance characteristics of the sample battery are determined to construct a sample battery feature set, and the sample battery training set is selected from the sample battery feature set.

[0121] For example, the number of cycles, charge and discharge current, charge and discharge voltage, and battery capacity are extracted from the single test data of each sample battery to produce a sample battery feature set, while excluding the influence of temperature, pressure, etc. in the single test data. When constructing a sample battery feature set with single test data, the charge and discharge current, charge and discharge voltage, battery capacity, number of cycles, etc. can be normalized or standardized to reduce the data dimension.

[0122] In 202, a regression model to be trained is obtained.

[0123] Among them, the machine learning model that is not trained using the sample battery feature set is called a regression model, and the machine learning model that is trained using the sample battery feature set is called a cycle performance prediction model.

[0124] The use of a regression model can utilize the battery performance characteristics corresponding to the known number of consecutive cycles to predict the battery cycle performance indicators under the later unknown number of cycles, thereby realizing the prediction of the battery cycle performance and facilitating the optimization of the battery cycle performance in the subsequent production and manufacturing process.

[0125] In some embodiments, 202 further includes:

[0126] 2021. Based on the network search algorithm, traverse the regression model to obtain multiple candidate hyperparameter combinations.

[0127] The grid search algorithm (GridSearchCV) is a hyperparameter optimization method. By setting a series of possible values ​​for each hyperparameter, and then trying all possible combinations to obtain multiple candidate hyperparameter combinations.

[0128] In this embodiment, by using a network search algorithm to clarify various candidate hyperparameter combinations that can be used by the regression model, and then intuitively and parallelly evaluating various candidate hyperparameter combinations, the efficiency of optimizing the hyperparameters of the regression model is improved, and the speed of model training is improved. Furthermore, different hyperparameters can be flexibly combined, which improves the ability to understand the hyperparameters of the regression model and helps improve the model prediction performance.

[0129] 2022. A cross-validation algorithm is used to divide the sample battery data set into a sample battery training set and a sample battery verification set, and the performance of the regression model under each candidate hyperparameter combination is verified to obtain the performance verification results corresponding to each candidate hyperparameter.

[0130] The sample battery data set is divided into a sample battery training set and a sample battery validation set each time through the cross-validation algorithm. The regression model is trained using the sample battery training set, and the performance of the regression model is evaluated on the sample battery validation set. Then, the optimal hyperparameter combination is determined based on the performance verification results.

[0131] In this embodiment, the cross-validation algorithm can adopt 3-fold cross-validation. When in use, the sample battery data set can be first randomly divided into three equal subsets. In the first iteration, the first subset is used as the sample battery training set, and the second and third subsets are combined as the sample battery verification set. In the second iteration, the second subset is used as the sample battery training set, and the first and third subsets are combined as the sample battery verification set. In the third iteration, the third subset is used as the sample battery training set, and the first and second subsets are combined as the sample battery verification set.

[0132] This embodiment uses a cross-validation method to adjust the hyperparameters of the regression model according to the performance verification results. For example, when the structure of the regression model is similar to a random forest machine learning model, the number of trees, the number of features considered when splitting each node, etc. can be adjusted, and the optimization of the hyperparameters can be repeated continuously. In addition, this method of dividing the training set and the validation set reduces the subjectivity and contingency of data division, avoids deviations caused by unreasonable data division, and can maximize the use of information in the data set, thereby improving data utilization. In addition, more comprehensive data utilization can be used to more accurately evaluate model performance, and then the best target hyperparameter combination can be selected to improve the stability and generalization ability of the regression model.

[0133] 2023. According to the performance verification results, determine the target hyperparameter combination from multiple candidate hyperparameter combinations.

[0134] The parameter values ​​in the candidate hyperparameter combination are adjusted based on the performance verification results to obtain the target hyperparameter combination with the best performance.

[0135] In this embodiment, the target hyperparameter combination of the regression model is determined by combining the network search algorithm and the cross-validation algorithm, which not only improves the feature selection ability of the model, but also enhances the generalization ability of the model. In addition, the regression model also takes into account the correlation between multiple features and the influence of time factors, can adapt to the processing of multi-dimensional features, and has good prediction accuracy and stability.

[0136] 2024. Construct a regression model to be trained based on the target hyperparameter combination.

[0137] Among them, the specific structure of the regression model can be determined according to the target hyperparameter combination, such as the number of trees, maximum depth, minimum number of sample splits, etc. Then, a large number of sample feature data sets are used to train it again, which can effectively improve the subsequent prediction performance of the model.

[0138] In 203, the regression model is trained based on the sample battery training set until a preset condition is met, thereby obtaining a trained cycle performance prediction model.

[0139] The sample battery training set may be larger or more recent than the sample battery training set used in the cross validation. The preset conditions may include using the sample battery training set to train the regression model until a preset number of training times is reached, or training until the difference between the predicted value and the true value of the regression model is minimized, etc., which may be set according to actual needs.

[0140] In some embodiments, 203 further includes:

[0141] Obtaining the performance characteristics of the regression model to be trained, including battery capacity, number of cycles, charge and discharge voltage, and charge and discharge current;

[0142] The number of cycles, charge and discharge voltage, and charge and discharge current in each single test cycle in the sample battery training set are used as the input of the regression model, and the battery capacity in each single test cycle is used as the output of the regression model. The regression model is trained until the preset conditions are met to obtain a cycle performance prediction model.

[0143] When training the regression model, the charge and discharge voltage and charge and discharge current corresponding to each cycle number are used as model input, and the battery capacity corresponding to the cycle number is used as model output. By learning the corresponding relationship between the model input and the model output, a trained cycle performance prediction model is obtained.

[0144] In some embodiments, during the training of the regression model, the method further comprises:

[0145] Using an evaluation index to evaluate the model performance of the cycle performance prediction model, the evaluation index includes at least one of a mean square error, a root mean square error, and a determination coefficient; or,

[0146] By displaying a scatter plot of the predicted values ​​and true values ​​of the cycle performance prediction model, the model performance of the cycle performance prediction model can be evaluated based on the scatter plot.

[0147] Specifically, the model performance of the cycle performance prediction model can also be evaluated. This embodiment provides two methods. When using evaluation indicators for evaluation, any indicator can be used, as follows:

[0148] The mean square error (MSE) is to first calculate the difference between the predicted value of the cycle performance prediction model and the true value, and then calculate the average value of the sum of the squares of the differences. Among them, the smaller the mean square error, the smaller the prediction error of the cycle performance prediction model, and the larger the mean square error, the larger the prediction error of the cycle performance prediction model. When the mean square error is small, it means that the predicted value of the cycle performance prediction model is closer to the actual value, and the prediction ability of the model is stronger.

[0149] Root mean square error (RMSE) is the square root of the mean square error, and is also used to measure the error size of the cycle performance prediction model. Among them, the root mean square error is easier to understand and explain than the mean square error. And the smaller the value of the root mean square error, the higher the prediction accuracy of the model.

[0150] Coefficient of determination (R 2 ) is a statistical indicator used to evaluate the goodness of fit of the cycle performance prediction model to the data. 2The value of is between 0 and 1. The closer it is to 1, the better the fit of the cycle performance prediction model to the data, that is, the higher the degree of data variation that the model can explain. 2 A value of 1 indicates that the cycle performance prediction model fits the data perfectly, while R 2 A value of 0 means that the cycle performance prediction model cannot explain the variation in the data.

[0151] In this embodiment, the model performance of the cycle performance prediction model is evaluated by using evaluation indicators, which helps to quantify the accuracy of the cycle performance prediction model and the degree of fit of the cycle performance prediction model to the data. On this basis, it is convenient to perform intuitive performance evaluation or model tuning of the cycle performance prediction model.

[0152] Specifically, when the model performance of the cycle performance prediction model is poor according to the evaluation index, the above evaluation index can be narrowed as a target to adjust the parameters of the cycle performance prediction model, thereby accurately improving the prediction performance of the model.

[0153] When evaluating the performance of the cycle performance prediction model by displaying a scatter plot, by marking the corresponding positions of the actual value and the predicted value in the chart, it is convenient to better understand the distribution and subsequent trends of the two data, and then intuitively understand the prediction effect of the cycle performance prediction model, as well as the deviation between the actual value and the predicted value. For example, if most of the points are concentrated near the trend line, it means that the prediction of the cycle performance prediction model is relatively accurate; if the points are scattered on both sides of the trend line, it means that the prediction error of the cycle performance prediction model is large.

[0154] After being able to intuitively evaluate the relationship between the distribution of scatter points and the trend line, the model parameters of the cycle performance prediction model can be further adjusted so that the scattered positions of the scatter points fit the trend line to the greatest extent possible, thereby improving the model performance.

[0155] See also Figure 4 , Figure 4 The schematic diagram of the structure of the device for predicting battery cycle performance provided in the embodiment of the present application is as follows: The device 300 for predicting battery cycle performance may include: a feature acquisition module 301 , a model acquisition module 303 , and an index prediction module 302 .

[0156] The feature acquisition module 301 is used to acquire the performance features of the battery under test from the previous usage data of the battery under test;

[0157] The index prediction module 302 is used to predict the performance characteristics based on the pre-trained cycle performance prediction model to obtain the cycle performance index of the battery to be tested.

[0158] In one implementation, the feature acquisition module 301 is further configured to:

[0159] Obtaining the previous usage data of the battery to be tested, the previous usage data including the charge and discharge voltage data, charge and discharge current data, battery capacity data, temperature data and pressure data at each historical cycle number;

[0160] Perform data cleaning on the previously used data to obtain processed previously used data;

[0161] The performance characteristics of the battery to be tested are determined based on the charge and discharge voltage data, charge and discharge current data, and battery capacity data at each historical cycle number in the processed previous usage data.

[0162] In one embodiment, the cycle performance prediction model includes a feature sampling layer, a feature classifier, an estimator, and a result output layer connected in sequence, the feature classifier includes a plurality of parallel sub-feature classifiers; the performance characteristics include the historical number of cycles, the historical charge and discharge voltage, and the historical charge and discharge current, and the cycle performance index is measured by the battery capacity; the index prediction module 302 is also used for:

[0163] Based on the feature sampling layer, historical charge and discharge currents and historical charge and discharge voltages at multiple different historical cycle numbers are selected from the performance features as inputs of the feature classifier;

[0164] Each sub-feature classifier performs feature classification processing on a historical charge and discharge current and a historical charge and discharge voltage under a historical cycle number inputted by each sub-feature classifier, and obtains a feature classification result of each sub-feature classifier;

[0165] The classification results of each feature are summarized and processed by the estimator to obtain the estimation result;

[0166] The estimated results are predicted through the result output layer to obtain the target battery capacity under the target number of cycles.

[0167] In one embodiment, the indicator prediction module 302 is further used to:

[0168] The estimated results are predicted through the result output layer to obtain the predicted battery capacity under the target cycle number;

[0169] Count the average temperature of temperature data under each historical cycle number;

[0170] The predicted battery capacity is adjusted according to the temperature average value to obtain the target battery capacity.

[0171] In one embodiment, the indicator prediction module 302 is further used to:

[0172] Determine the number weight of the historical number of cycles corresponding to each sub-feature classifier through an estimator;

[0173] The feature classification results and frequency weights corresponding to each sub-feature classifier are weighted by the estimator to obtain the weighted classification results corresponding to each sub-feature classifier;

[0174] The weighted classification results are aggregated through the estimator to obtain the estimated result.

[0175] See also Figure 5 , Figure 5 Another structural schematic diagram of the device for predicting battery cycle performance provided in an embodiment of the present application. The device 300 for predicting battery cycle performance also includes a model acquisition module 303, which is communicatively connected to a model training module 304, and the model training module 304 can be integrated in the model training device, and can also be integrated in the device 300 for predicting battery cycle performance.

[0176] In one embodiment, the model training module 304 is used to:

[0177] Obtaining a sample battery training set, the sample battery training set includes a plurality of cycle test feature sets of sample batteries, and the cycle test feature set of each sample battery includes its own performance characteristics of the sample battery tested in each single test cycle;

[0178] Get the regression model to be trained;

[0179] The regression model is trained based on the sample battery training set until the preset conditions are met, and a trained cycle performance prediction model is obtained.

[0180] In one embodiment, the model training module 304 is further used to:

[0181] Based on the network search algorithm, the regression model is traversed to obtain multiple candidate hyperparameter combinations;

[0182] A cross-validation algorithm is used to divide the sample battery data set into a sample battery training set and a sample battery validation set, and the performance of the regression model under each candidate hyperparameter combination is verified to obtain the performance verification result corresponding to each candidate hyperparameter;

[0183] According to the performance verification results, a target hyperparameter combination is determined from multiple candidate hyperparameter combinations;

[0184] Construct a regression model to be trained based on the target hyperparameter combination.

[0185] In one embodiment, the performance characteristics of the regression model to be trained include battery capacity, cycle number, charge and discharge voltage, and charge and discharge current; the model training module 304 is also used to:

[0186] The number of cycles, charge and discharge voltage, and charge and discharge current in each single test cycle in the sample battery training set are used as the input of the regression model, and the battery capacity in each single test cycle is used as the output of the regression model. The regression model is trained until the preset conditions are met to obtain a cycle performance prediction model.

[0187] In one embodiment, during the training of the regression model, the model training module 304 is further used to:

[0188] Using an evaluation index to evaluate the model performance of the cycle performance prediction model, the evaluation index includes at least one of a mean square error, a root mean square error, and a determination coefficient; or,

[0189] By displaying a scatter plot of the predicted values ​​and true values ​​of the cycle performance prediction model, the model performance of the cycle performance prediction model can be evaluated based on the scatter plot.

[0190] In one embodiment, the model training module 304 is further used to:

[0191] Respectively obtain cycle test data of each sample battery within a preset period, wherein each sample battery has a set of single test data in each single test cycle;

[0192] Performing data cleaning processing on the cycle test data of each sample battery to obtain processed cycle test data;

[0193] The performance characteristics of the sample battery are determined from the processed cycle test data to construct a sample battery feature set, and the sample battery training set is selected from the sample battery feature set.

[0194] See also Figure 6 , Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices.

[0195] An embodiment of the present application also provides an electronic device, the electronic device 400 includes a memory 401 and a processor 402, and the processor 402 is used to execute the process in the method for predicting battery cycle performance provided in this embodiment by calling a computer program stored in the memory 401.

[0196] Those skilled in the art will understand that Figure 6The electronic device structure shown in the figure does not constitute a limitation of the electronic device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0197] The application programs stored in the memory 401 include computer programs. The processor 402 executes various functional applications and data processing by running the computer programs stored in the memory 401 .

[0198] The processor 402 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. It executes various functions of the electronic device and processes data by running or executing computer programs stored in the memory 401 and calling data stored in the memory 401, thereby monitoring the electronic device as a whole.

[0199] In this embodiment, the processor 402 in the electronic device will load the executable code corresponding to the process of one or more application programs into the memory 401 according to the following instructions, and the processor 402 will run the application program stored in the memory 401 to execute:

[0200] Obtaining performance characteristics of the battery under test from previous usage data of the battery under test;

[0201] The performance characteristics are predicted based on the pre-trained cycle performance prediction model to obtain the cycle performance index of the battery to be tested.

[0202] In one embodiment, in obtaining the performance characteristics of the battery to be tested from the previous usage data of the battery to be tested, the processor 402 executes: obtaining the previous usage data of the battery to be tested, the previous usage data including the charge and discharge voltage data, charge and discharge current data, battery capacity data, temperature data and pressure data at each historical cycle number; performing data cleaning processing on the previous usage data to obtain processed previous usage data; and determining the performance characteristics of the battery to be tested based on the charge and discharge voltage data, charge and discharge current data and battery capacity data at each historical cycle number in the processed previous usage data.

[0203] In one embodiment, in predicting and processing the estimated results through the result output layer to obtain the target battery capacity under the target number of cycles, the processor 402 executes: predicting and processing the estimated results through the result output layer to obtain the predicted battery capacity under the target number of cycles; calculating the temperature average value of the temperature data under each historical number of cycles; and adjusting the predicted battery capacity according to the temperature average value to obtain the target battery capacity.

[0204] In one embodiment, in aggregating the feature classification results through an estimator to obtain an estimated result, the processor 402 executes: determining the number weights of the historical number of cycles corresponding to each sub-feature classifier through an estimator; weighting the feature classification results and number weights corresponding to each sub-feature classifier through an estimator to obtain a weighted classification result corresponding to each sub-feature classifier; and aggregating the weighted classification results through an estimator to obtain an estimated result.

[0205] In one embodiment, before obtaining a pre-trained cycle performance prediction model, the processor 502 executes obtaining a sample battery training set, the sample battery training set includes a cycle test feature set of multiple sample batteries, and the cycle test feature set of each sample battery includes the inherent performance characteristics of the sample battery measured under each single test cycle; obtaining a regression model to be trained; and training the regression model based on the sample battery training set until a preset condition is met to obtain a trained cycle performance prediction model.

[0206] In one embodiment, in obtaining the regression model to be trained, the processor 502 executes a network search algorithm to traverse the regression model to obtain multiple candidate hyperparameter combinations; a cross-validation algorithm is used to divide the sample battery data set into a sample battery training set and a sample battery verification set, and the performance of the regression model under each candidate hyperparameter combination is verified to obtain a performance verification result corresponding to each candidate hyperparameter; based on the performance verification result, a target hyperparameter combination is determined from multiple candidate hyperparameter combinations; and the regression model to be trained is constructed based on the target hyperparameter combination.

[0207] In one embodiment, the regression model is trained based on a sample battery training set until a preset condition is met to obtain a trained cycle performance prediction model, and the processor 502 uses the number of cycles, charge and discharge voltage, and charge and discharge current under each single test cycle in the sample battery training set as the input of the regression model, and uses the battery capacity under each single test cycle as the output of the regression model, and trains the regression model until the preset condition is met to obtain a cycle performance prediction model, wherein the performance characteristics of the regression model to be trained include battery capacity, number of cycles, charge and discharge voltage, and charge and discharge current.

[0208] In one embodiment, during the training of the regression model, the processor 502 performs: evaluating the model performance of the cycle performance prediction model using an evaluation indicator, the evaluation indicator including at least one of a mean square error, a root mean square error, and a coefficient of determination; or, evaluating the model performance of the cycle performance prediction model based on a scatter plot by displaying a scatter plot of the predicted value and the true value of the cycle performance prediction model.

[0209] In one embodiment, before obtaining the sample battery training set, the processor 502 executes: respectively obtaining the cycle test data of each sample battery within a preset period, wherein each sample battery has a set of single test data under each single test cycle; performing data cleaning processing on the cycle test data of each sample battery to obtain processed cycle test data; determining the performance characteristics of the sample battery itself from the processed cycle test data to construct a sample battery feature set, and the sample battery training set is selected from the sample battery feature set.

[0210] In one embodiment, in predicting the performance characteristics based on the pre-trained cycle performance prediction model to obtain the cycle performance index of the battery to be tested, the processor 502 executes: based on the feature sampling layer, selects the historical charge and discharge current and historical charge and discharge voltage under multiple different historical cycle numbers from the performance characteristics as the input of the feature classifier; performs feature classification processing on the historical charge and discharge current and historical charge and discharge voltage under a historical cycle number input by each sub-feature classifier to obtain the feature classification results of each sub-feature classifier; summarizes and processes each feature classification result through the estimator to obtain the estimation result; predicts and processes the estimation result through the result output layer to obtain the target battery capacity under the target cycle number. Among them, the cycle performance prediction model includes a feature sampling layer, a feature classifier, an estimator and a result output layer connected in sequence, and the feature classifier includes multiple parallel sub-feature classifiers; the performance characteristics include the historical cycle number, the historical charge and discharge voltage and the historical charge and discharge current, and the cycle performance index is measured by the battery capacity.

[0211] In the above embodiments, the description of each embodiment has its own emphasis. For the part that is not described in detail in a certain embodiment, please refer to the detailed description of the method for predicting battery cycle performance above, which will not be repeated here.

[0212] The electronic device provided in the embodiment of the present application belongs to the same concept as the method for predicting battery cycle performance in the above embodiment. Any method provided in the method for predicting battery cycle performance embodiment can be run on the electronic device. The specific implementation process is detailed in the method for predicting battery cycle performance embodiment, which will not be repeated here.

[0213] It should be noted that, for the method for predicting battery cycle performance in the embodiment of the present application, ordinary technicians in this field can understand that all or part of the process of implementing the method for predicting battery cycle performance in the embodiment of the present application can be completed by controlling the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, such as stored in a memory, and executed by at least one processor, and the execution process may include the process of the embodiment of the method for predicting battery cycle performance. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM, Read Only Memory), a random access memory (RAM, Random Access Memory), etc.

[0214] An embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed on a computer, the computer executes the process in the method for predicting battery cycle performance provided in the present embodiment.

[0215] For the electronic device of the embodiment of the present application, each functional module thereof may be integrated into a processing chip, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk.

[0216] The above is a detailed introduction to the method, device and storage medium for predicting battery cycle performance provided in the embodiments of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the ideas of the present application, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for predicting battery cycle performance, characterized in that: The method comprises: Obtaining performance characteristics of the battery under test from previous usage data of the battery under test; The performance characteristics are predicted based on a pre-trained cycle performance prediction model to obtain a cycle performance index of the battery to be tested.

2. The method according to claim 1, characterized in that The step of obtaining the performance characteristics of the battery to be tested from the previous usage data of the battery to be tested includes: Acquire the previous usage data of the battery to be tested, wherein the previous usage data includes charge and discharge voltage data, charge and discharge current data, battery capacity data, temperature data and pressure data at each historical cycle number; Performing data cleaning processing on the previously used data to obtain processed previously used data; The performance characteristics of the battery to be tested are determined according to the charge and discharge voltage data, charge and discharge current data and battery capacity data at each historical cycle number in the processed previous usage data.

3. The method according to claim 2, characterized in that The cycle performance prediction model includes a feature sampling layer, a feature classifier, an estimator and a result output layer connected in sequence, and the feature classifier includes a plurality of parallel sub-feature classifiers; the performance characteristics include the historical number of cycles, the historical charge and discharge voltage and the historical charge and discharge current, and the cycle performance index is measured by the battery capacity; The performance characteristics are predicted based on the pre-trained cycle performance prediction model to obtain the cycle performance index of the battery to be tested, including: Selecting historical charge and discharge currents and historical charge and discharge voltages at multiple different historical cycle numbers from the performance characteristics based on the feature sampling layer as inputs to the feature classifier; Performing feature classification processing on the historical charge and discharge current and the historical charge and discharge voltage at a historical cycle number input by each of the sub-feature classifiers, thereby obtaining feature classification results of each of the sub-feature classifiers; The estimator aggregates the feature classification results to obtain an estimation result; The estimation result is predicted and processed by the result output layer to obtain the target battery capacity under the target cycle number.

4. The method according to claim 3, characterized in that: The predicting process of the estimation result by the result output layer to obtain the target battery capacity under the target cycle number includes: Performing prediction processing on the estimation result through the result output layer to obtain the predicted battery capacity under the target cycle number; Counting the average temperature of the temperature data at each of the historical cycle times; The predicted battery capacity is adjusted according to the temperature average value to obtain a target battery capacity.

5. The method according to claim 3, characterized in that: The estimator summarizes and processes the feature classification results to obtain an estimation result, including: Determine the number weight of the historical number of cycles corresponding to each of the sub-feature classifiers by an estimator; The estimator performs weighted processing on the feature classification results and frequency weights corresponding to each of the sub-feature classifiers to obtain weighted classification results corresponding to each of the sub-feature classifiers; The estimator aggregates the weighted classification results to obtain an estimation result.

6. The method according to any one of claims 1 to 5, characterized in that: The cycle performance prediction model is obtained by training a regression model with a sample battery training set, wherein the sample battery training set includes a cycle test feature set of multiple sample batteries, and the cycle test feature set of each sample battery includes the battery capacity, cycle number, charge and discharge voltage, and charge and discharge current of the sample battery measured in each single test cycle; When training the regression model, the number of cycles, charge and discharge voltage, and charge and discharge current under each single test cycle in the sample battery training set are used as the input of the regression model, and the battery capacity under each single test cycle is used as the output of the regression model. The regression model is trained until the preset conditions are met to obtain the cycle performance prediction model.

7. The method according to claim 6, characterized in that The regression model is constructed according to a target hyperparameter combination, and a method for obtaining the target hyperparameter combination includes: Based on the network search algorithm, the regression model is traversed to obtain multiple candidate hyperparameter combinations; A cross-validation algorithm is used to divide the sample battery data set into a sample battery training set and a sample battery verification set, and the performance of the regression model under each candidate hyperparameter combination is verified to obtain a performance verification result corresponding to each candidate hyperparameter; According to the performance verification result, a target hyperparameter combination is determined from the plurality of candidate hyperparameter combinations.

8. The method according to claim 6, characterized in that In the process of training the regression model, the method further includes: Using an evaluation index to evaluate the model performance of the cycle performance prediction model, the evaluation index includes at least one of a mean square error, a root mean square error, and a determination coefficient; or, By displaying a scatter plot of the predicted value and the true value of the cycle performance prediction model, the model performance of the cycle performance prediction model is evaluated based on the scatter plot.

9. An electronic device, characterized in that: include: processor; as well as A memory storing a computer program, which, when executed by the processor, enables the processor to perform the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the method according to any one of claims 1 to 7.

Images