Test model for battery performance test, battery performance test method, device, equipment, medium and product

Abstract

The invention discloses a test model for battery performance test, a battery performance test method, an electronic device, electronic equipment, a computer readable storage medium and a computer program product, and the test model comprises: training a preset random forest model according to training data determined by battery sample data, and determining the test model. Therefore, the performance test result of the to-be-tested battery can be determined based on the test model determined by the training random forest model, so that the situation that the performance test result of the to-be-tested battery needs to be obtained by performing a real battery performance test on the to-be-tested battery, such as an intermittent cycle test with a relatively long test period, is avoided; therefore, the proportion of the battery performance test link in the whole battery development process can be shortened, the battery development cost such as time cost is reduced, and the battery development efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a test model for battery performance testing, a battery performance testing method, an electronic device, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] In related technologies, intermittent cycle testing can be performed to evaluate the performance and lifespan of lithium-ion batteries. This allows researchers to study their performance under different usage conditions, including capacity retention, charge and discharge efficiency, and cycle life. However, intermittent cycle testing requires a long time, typically 30 days or more, resulting in a longer battery development cycle, which in turn affects development efficiency and costs. Summary of the Invention

[0003] The present application provides a test model for battery performance testing, a battery performance testing method, an electronic device, an electronic device, a computer-readable storage medium, and a computer program product.

[0004] The present application provides a test model for battery performance testing, including:

[0005] The preset random forest model is trained based on the training data determined by the battery sample data to determine the test model.

[0006] Thus, in the embodiment of the present application, a pre-set random forest model can be trained to determine a test model suitable for battery performance testing. Then, the performance test results of the battery to be tested can be determined based on the test model determined by the trained random forest model, thereby avoiding the need to conduct a real battery performance test on the battery to be tested to obtain the performance test results, such as intermittent cycle testing with a long test cycle, etc., thereby shortening the proportion of the battery performance test link in the entire battery development process to a certain extent, reducing battery development costs such as time costs, and improving battery development efficiency. In addition, because the test model is obtained by training based on the random forest model, the data required for model training is less than that for neural network models such as convolutional neural network models, the training difficulty is lower and the training cost is lower, making it more suitable for battery performance testing scenarios with higher real-time requirements.

[0007] In certain embodiments of the present application, the battery sample data includes at least one of chemical and cell parameters of the battery sample and performance test results of the battery sample.

[0008] Thus, in the embodiment of the present application, the random forest model can be trained using at least one of the chemical and cell parameters of the battery sample and the performance test results of the battery sample, thereby ensuring the robust execution of the model training process and the model performance to a certain extent.

[0009] In certain embodiments of the present application, the chemical and cell parameters include cell thickness, cell width, cell height, cut-off voltage, positive electrode type, positive electrode plate composition data, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode plate composition data, negative electrode loading, negative electrode compaction, negative-to-positive electrode capacity ratio, electrolyte residual amount, electrolyte composition data and at least one of the diaphragm category.

[0010] Thus, in an embodiment of the present application, model training can be performed based on at least one of the cell thickness, cell width, cell height, cut-off voltage, positive electrode type, positive electrode plate composition data, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode plate composition data, negative electrode loading, negative electrode compaction, negative-to-positive electrode capacity ratio, electrolyte residual amount, electrolyte composition data, and diaphragm category of the battery sample, thereby ensuring the prediction accuracy of the model.

[0011] In certain embodiments of the present application, the performance test results include intermittent cycle test results of the battery sample.

[0012] Thus, in the embodiment of the present application, a random forest model suitable for the task of predicting battery intermittent cycle test results can be trained through the chemical and cell parameters of the battery samples and the intermittent cycle test results of the battery samples, that is, a test model, thereby avoiding the need to conduct real intermittent cycle tests on the battery to be tested to obtain the intermittent cycle test results of the battery to be tested, thereby reducing the proportion of intermittent cycle tests in the battery performance test link to a certain extent, and thereby reducing the time cost of the battery test link, thereby improving the battery test efficiency.

[0013] In certain embodiments of the present application, the training data determined based on the battery sample data includes:

[0014] Performing preprocessing on the battery sample data to determine preprocessed battery sample data;

[0015] The training data is determined according to the preprocessed battery sample data.

[0016] Thus, in an embodiment of the present application, the battery sample data can be preprocessed to determine the preprocessed battery sample data, and the training data can be determined based on the preprocessed battery sample data, so that the training data can be determined based on the preprocessed battery sample data, thereby ensuring the effectiveness and reliability of the training data to a certain extent, and further ensuring the performance of the random forest model trained based on the training data.

[0017] In certain embodiments of the present application, the preprocessing includes at least one of blank field filling, preset field conversion, and preset field deletion.

[0018] Thus, in the embodiment of the present application, at least one of blank field filling, preset field conversion, and preset field deletion may be performed on the battery sample data, thereby completing the preprocessing of the battery sample data.

[0019] In certain embodiments of the present application, determining the training data based on the pre-processed battery sample data includes:

[0020] Determine a plurality of groups of pre-processed battery sample data, and determine the training data according to the plurality of groups of pre-processed battery sample data.

[0021] Thus, in the embodiment of the present application, multiple groups of the pre-processed battery sample data can be determined, and the training data can be determined based on the multiple groups of the pre-processed battery sample data, thereby achieving the determination of the training data and ensuring the quality of the training data to a certain extent.

[0022] In certain embodiments of the present application, determining the training data based on multiple sets of pre-processed battery sample data includes:

[0023] Determine a portion of the preprocessed battery sample data as a training set, and determine a portion of the preprocessed battery sample data as a test set, wherein the preprocessed battery sample data included in the training set and the preprocessed battery sample data included in the test set are at least partially staggered, and determine the training data based on the test set and the training set.

[0024] Thus, in the embodiment of the present application, part of the group of preprocessed battery sample data can be determined as a training set, and part of the group of preprocessed battery sample data can be determined as a test set, thereby completing the construction of the training data and ensuring the robustness of subsequent model training steps.

[0025] In certain embodiments of the present application, determining a portion of the preprocessed battery sample data as a training set and determining a portion of the preprocessed battery sample data as a test set includes:

[0026] Perform multiple preset processing on multiple groups of the preprocessed battery sample data to determine multiple sets of the training data, wherein the preset processing includes: determining some groups of the preprocessed battery sample data as training sets, and determining some groups of the preprocessed battery sample data as test sets, the preprocessed battery sample data included in the training sets and the preprocessed battery sample data included in the test sets are at least partially staggered, and one set of the training data does not completely overlap with another set of the training data.

[0027] Thus, in the embodiment of the present application, multiple sets of pre-processed battery sample data may be subjected to multiple preset processes to determine multiple sets of training data, thereby achieving the construction of training data.

[0028] In certain embodiments of the present application, the random forest model includes multiple models, the training data includes a training set and a test set, and the training data determined according to the battery sample data is used to train the preset random forest model to determine the test model, including:

[0029] Training the random forest model according to the training set to determine a trained random forest model;

[0030] Determining model performance data of the trained random forest model based on the test set and the trained random forest model;

[0031] At least one of the trained random forest models is determined as the test model based on the model performance data.

[0032] Thus, in an embodiment of the present application, the random forest model can be trained based on the training set to determine the trained random forest model, and the model performance data of the trained random forest model can be determined based on the test set and the trained random forest model, and at least one of the multiple trained random forest models can be determined as the test model based on the model performance data of each trained random forest model, thereby completing the training and testing of the test model, so that the test model can be robustly determined.

[0033] In certain embodiments of the present application, the model performance data includes recognition accuracy, and determining at least one of the trained random forest models as the test model based on the model performance data includes:

[0034] The trained random forest model with the highest recognition accuracy among the trained random forest models is determined as the test model.

[0035] Thus, in the embodiment of the present application, the trained random forest model with the highest recognition accuracy among multiple trained random forest models can be determined as the test model, thereby ensuring that the test model can effectively and reliably predict the battery performance test results.

[0036] In certain embodiments of the present application, the battery sample data includes battery parameters and performance test results, and determining the model performance data of the trained random forest model based on the test set and the trained random forest model includes:

[0037] Determining a performance prediction result of the trained random forest model for the battery sample based on the trained random forest model and the battery parameters of the battery sample;

[0038] The model performance data is determined based on the performance prediction results of the trained random forest model for each of the battery samples and the performance test results of each of the battery samples.

[0039] Thus, in the embodiment of the present application, the performance prediction results of the trained random forest model for the battery sample can be determined based on the trained random forest model and the battery parameters of the battery sample, and the model performance data can be determined based on the performance prediction results of the trained random forest model for each battery sample and the performance test results of each battery sample, thereby achieving the determination of the model performance data, and then the test model can be robustly determined based on the model performance data.

[0040] In certain embodiments of the present application, a preset hyperparameter is adjusted to determine at least one adjusted hyperparameter;

[0041] Determine a plurality of random forest models based on at least one of the adjusted hyperparameters.

[0042] Thus, in an embodiment of the present application, the pre-set hyperparameters can be adjusted to determine at least one adjusted hyperparameter, and multiple random forest models can be determined based on at least one of the adjusted hyperparameters, thereby achieving the determination of multiple random forest models.

[0043] In certain embodiments of the present application, the hyperparameters include the number of decision trees and / or the depth of decision trees.

[0044] Thus, in the embodiment of the present application, the number of decision trees and / or the depth of decision trees can be preset to construct multiple random forest models to be trained.

[0045] The present application provides a battery performance testing method, including:

[0046] According to the battery data to be tested and the above-mentioned test model, a performance test result of the battery to be tested corresponding to the battery data to be tested is determined.

[0047] An embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the test model for battery performance testing or the battery performance testing method is implemented.

[0048] An embodiment of the present application provides an electronic device, including the above-mentioned electronic device.

[0049] An embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, it implements the above-mentioned test model for battery performance testing, or implements the above-mentioned battery performance testing method.

[0050] An embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-mentioned test model for battery performance testing, or implements the above-mentioned battery performance testing method.

[0051] The battery performance testing method, electronic device, electronic device, computer-readable storage medium, and computer program product provided by the embodiments of the present application can train a pre-set random forest model to determine a test model suitable for battery performance testing. Furthermore, the performance test results of the battery to be tested can be determined based on the test model determined by the trained random forest model, thereby avoiding the need to conduct a real battery performance test on the battery to be tested to obtain the performance test results of the battery to be tested, such as an intermittent cycle test with a long test cycle, etc., thereby shortening the proportion of the battery performance test link in the entire battery development process to a certain extent, reducing battery development costs such as time costs, and improving battery development efficiency. In addition, because the test model is obtained through training based on the random forest model, the data required for model training is less than that for neural network models such as convolutional neural network models, the training difficulty is lower and the training cost is lower, which is more suitable for battery performance testing scenarios with higher real-time requirements.

[0052] Additional aspects and advantages of the embodiments of the present application will be given in part in the description below, and in part will become obvious from the description below, or will be learned through practice of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0054] Figure 1 A flow chart of a test model for battery performance testing in certain embodiments of the present application;

[0055] Figure 2 A flow chart of a test model for battery performance testing in certain embodiments of the present application;

[0056] Figure 3 A flow chart of a test model for battery performance testing in certain embodiments of the present application;

[0057] Figure 4 A flow chart of a test model for battery performance testing in certain embodiments of the present application;

[0058] Figure 5 A flow chart of a test model for battery performance testing in certain embodiments of the present application;

[0059] Figure 6 This is a schematic diagram of an application scenario in some embodiments of the present application;

[0060] Figure 7 This is a schematic diagram of an application scenario in some embodiments of the present application;

[0061] Figure 8 This is a schematic diagram of an application scenario in some embodiments of the present application;

[0062] Figure 9 Schematic diagram of the process of battery performance testing method in certain embodiments of the present application. DETAILED DESCRIPTION

[0063] The embodiments of the present application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the embodiments of the present application, and should not be understood as limiting the embodiments of the present application.

[0064] In related technologies, there are two main methods for predicting battery life: model analysis and data analysis. The model analysis method includes analysis based on electrochemical models and analysis based on equivalent circuit models.

[0065] Specifically, in the analysis based on the electrochemical model, the failure mechanism of the battery is mainly described from the perspectives of the internal ion diffusion of the battery, the Ohmic effect, the electrochemical kinetics, etc. This method has high accuracy, but it requires solving complex partial differential equations, which is computationally intensive. At the same time, the complex aging mechanism of the battery limits the application of this method to a certain extent. For example, in a scheme for predicting the cycle life of lithium batteries based on an electrochemical-thermal coupling model, the physical properties and electrochemical parameters of the lithium battery are first obtained, and then these parameters are used to establish an electrochemical-thermal coupling model, calculate the partial differential equations, and then verify the validity of the model. Then, the empirical life function is determined, and finally the parameters are fitted and solved to obtain the final life function. This scheme has the advantages of fast response, strong predictive ability, and a wide range of applications, but the electrochemical mechanism involved in this scheme is complex and requires solving a large number of partial differential equations, which limits the application of this scheme to a certain extent.

[0066] In contrast, in the analysis based on the equivalent circuit model, the electrical characteristic parameters of the battery, such as output voltage, SOC (State Of Charge, remaining capacity) and other parameters, can be used for prediction. Among them, the electrical characteristic parameters are easy to obtain and easy to calculate. However, the electrical characteristic parameters will continue to change with battery aging and differences in operating conditions, which will eventually affect the prediction accuracy. For example, in a future operating condition prediction scheme for a lithium-ion battery pack for space use, an equivalent model of the battery is first established and then the data is sparsely sampled. Then, parameter identification is performed based on an artificial immune algorithm to obtain an equivalent model after identification. Finally, according to the equivalent model, the current data under future operating conditions is injected into the model to predict the voltage change of the battery pack. It can be understood that the scope of application of this scheme is limited, and it is difficult to apply to life prediction that is significantly affected by battery aging and differences in operating conditions.

[0067] Furthermore, data analysis methods are based on machine learning, leveraging the nonlinear fitting capabilities of neural network models to extract deep features from various raw data for function fitting. However, function fitting for multi-dimensional raw data requires networks with large parameters and numerous layers. This results in large amounts of data required for network training, making it difficult to apply to development scenarios with high real-time requirements.

[0068] For example, in one battery remaining useful life prediction scheme, the cumulative discharge capacity [Q1, Q2, ..., QN] corresponding to a voltage point is first used as a model input feature to characterize the battery aging state. A convolutional neural network (CNN) is then used to extract battery aging information from the selected input features. A long short-term memory (LSTM) neural network then performs time series analysis on the CNN-extracted features, thereby constructing a battery remaining useful life prediction model. This scheme, by selecting features based on electrochemical background knowledge, avoids extracting features from raw voltage and current data, effectively reducing the size and complexity of the neural network. Furthermore, by extracting spatial correlation features from the data through the CNN and temporal features from these spatial correlation features through the LSTM, prediction accuracy can be effectively improved. However, this scheme requires the discharge curve of each battery cycle to extract input features. Batteries can cycle thousands of times. Recording and processing discharge data for each cycle would require extensive data manipulation and a high barrier to entry, making it difficult to widely apply and promote.

[0069] Based on the above problems you may encounter, please refer to Figure 1 , the embodiment of the present application provides a test model for battery performance testing, including:

[0070] 01: Train the pre-set random forest model based on the training data determined by the battery sample data to determine the test model.

[0071] Embodiments of the present application also provide an electronic device, comprising a memory and a processor. The test model for battery performance testing according to embodiments of the present application can be implemented by the electronic device according to embodiments of the present application. Specifically, the memory stores a computer program, and the processor is configured to train a pre-set random forest model based on training data determined from battery sample data to determine a test model.

[0072] Specifically, in an embodiment of the present application, an electronic device (or an electronic device in an electronic device) can obtain battery sample data, determine training data for model training, and then train a pre-set random forest model (Random Forest Model) based on the training data, thereby determining a random forest model suitable for battery performance testing, namely the above-mentioned test model.

[0073] Thus, in the embodiment of the present application, a pre-set random forest model can be trained to determine a test model suitable for battery performance testing. Then, the performance test results of the battery to be tested can be determined based on the test model determined by the trained random forest model, thereby avoiding the need to conduct a real battery performance test on the battery to be tested to obtain the performance test results, such as intermittent cycle testing with a long test cycle, etc., thereby shortening the proportion of the battery performance test link in the entire battery development process to a certain extent, reducing battery development costs such as time costs, and improving battery development efficiency. In addition, because the test model is obtained by training based on the random forest model, the data required for model training is less than that for neural network models such as convolutional neural network models, the training difficulty is lower and the training cost is lower, making it more suitable for battery performance testing scenarios with higher real-time requirements.

[0074] The random forest model is implemented using a bagging ensemble algorithm. Ensemble algorithms aggregate the modeling results of multiple estimators to produce a comprehensive result, thereby achieving better regression or classification performance than a single model. All base estimators in random forests are decision trees.

[0075] It is understandable that the forest composed of classification trees is called a random forest classifier, and the forest integrated by regression trees is called a random forest regressor.

[0076] It's also understandable that a random forest classifier is composed of multiple decision trees aggregated according to certain rules. Each decision tree is independent of the others, and each is trained based on randomly selected data. The higher the accuracy of a single decision tree, the higher the accuracy of the random forest classifier. The larger the number of decision trees, the better the model generally performs.

[0077] In one example, battery sample data can be understood as various types of data collected in advance, such as the battery cell thickness, cell width, electrolyte composition, and other battery chemical system and design data of the battery sample.

[0078] In one example, the training data can be determined by splitting and reorganizing the battery sample data. For example, the battery sample data is first divided into 10 groups, and then the first group is used as the test set and the second to tenth groups are used as training sets, thereby obtaining training data consisting of the test set and the training set.

[0079] In one example, when a random forest model is trained based on training data, the structure of each decision tree in the random forest model and the parameters that each decision tree depends on can be determined based on the training data.

[0080] In one example, the training process of a random forest model includes bootstrap sampling, decision tree construction, classification voting, and performance evaluation.

[0081] In the bootstrap sampling phase, T bootstrap samples are extracted from the training data with replacement to obtain the data subset D t .

[0082] In the decision tree construction process, each decision tree is based on random selection, starting from D t One or more data are selected to split at each node and thus a recursive tree structure is constructed. Therefore, any two decision trees can be relatively independent based on this random selection method.

[0083] It is understandable that the decision tree needs to find the best nodes and the best branching method, and the indicator for measuring this "best" is called "impurity". The larger the proportion of a certain type of label, the purer the leaves, the lower the impurity, and the better the branching.

[0084] It can also be understood that there are two indicators of impurity, namely Gini and Entropy.

[0085] Among them, Gini can be obtained by the following formula:

[0086]

[0087] Where p k is the proportion of samples belonging to category k.

[0088] Information entropy can be obtained by the following formula:

[0089]

[0090] Where K represents the number of label categories on the leaf node.

[0091] It is also understandable that when deciding whether to split a node, it is usually required that the difference between the parent node information entropy and the total information entropy of the child nodes be maximized, that is, the information gain is required to be maximized. The information gain can be calculated by the following formula, namely:

[0092]

[0093] Where D v It is a subset of D when the attribute A takes the value v.

[0094] In the classification voting phase, each decision tree in the model votes for one category, and the final predicted category is the category that receives the majority vote.

[0095] In performance evaluation, the performance of the model can be judged by one or more of the model's accuracy, precision, recall, F1-score, confusion matrix, ROC (Receiver Operating Characteristic Curve), and AUC (Area Under the Curve).

[0096] The calculation formula of accuracy is as follows:

[0097]

[0098] Where TP, TN, FP, and FN are the counts of true positives, true negatives, false positives, and false negatives, respectively.

[0099] The calculation formula of the accuracy is as follows:

[0100]

[0101] The calculation formula of recall rate is as follows:

[0102]

[0103] The calculation formula of F1-score is as follows:

[0104]

[0105] The confusion matrix is ​​shown in Table 1, namely:

[0106] Table 1

[0107] Predicted as positive Predicted as negative class Actually positive TP FN Actually negative FP TN

[0108] ROC refers to a curve with the false positive rate (FPR) at different thresholds as the horizontal axis and the recall rate (Recall) at different thresholds as the vertical axis. The calculation formula for the false positive rate is as follows:

[0109]

[0110] AUC is used to measure the overall performance. AUC represents the area under the ROC curve. The larger the area, the better the model performance.

[0111] In one example, the electronic device refers to a device that can be used to train a random forest model, such as a personal computer or server with sufficient computing power.

[0112] In certain embodiments of the present application, the battery sample data includes at least one of chemical and cell parameters of the battery sample and performance test results of the battery sample.

[0113] Specifically, in the embodiment of the present application, the battery sample data may include at least one of the chemical and cell parameters of the battery sample and the performance test results of the battery sample.

[0114] A battery sample can be understood as a battery that has undergone prior performance testing. The battery sample's chemical and cell parameters may include chemically relevant design parameters and cell specifications, such as electrolyte composition and cell dimensions.

[0115] In addition, the performance test results of the battery samples can be understood as the performance characterization of the battery samples obtained by performing performance tests on the battery samples. For example, for a binary performance test, the performance test results of the battery samples characterize whether the battery samples have passed the performance test. For another example, for a performance test for indicator measurement, the performance test results of the battery samples represent the specific value of a certain indicator of the battery sample.

[0116] It can be understood that the chemical and cell parameters of the battery sample may be related to the performance test results of the battery sample. Therefore, the embodiments of the present application can train the random forest model through the chemical and cell parameters of the battery sample, so that the random forest model can predict the performance test results of the battery sample based on the chemical and cell parameters of the battery sample.

[0117] Thus, in the embodiment of the present application, the random forest model can be trained using at least one of the chemical and cell parameters of the battery sample and the performance test results of the battery sample, thereby ensuring the robust execution of the model training process and the model performance to a certain extent.

[0118] In certain embodiments of the present application, the chemical and cell parameters include cell thickness, cell width, cell height, cut-off voltage, positive electrode type, positive electrode plate composition data, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode plate composition data, negative electrode loading, negative electrode compaction, negative-to-positive electrode capacity ratio, electrolyte residual amount, electrolyte composition data, and at least one of the diaphragm category.

[0119] Specifically, in an embodiment of the present application, a random forest model can be trained based on at least one of the cell thickness, cell width, cell height, cut-off voltage, positive electrode type, positive electrode sheet composition data, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode sheet composition data, negative electrode loading, negative electrode compaction, negative-to-positive electrode capacity ratio, electrolyte residual amount, electrolyte composition data, and diaphragm category of the battery sample.

[0120] In one example, the electronic device may train a random forest model based on the cell thickness, cell width, cell height, cut-off voltage, positive electrode type, positive electrode sheet composition data, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode sheet composition data, negative electrode loading, negative electrode compaction, negative-to-positive electrode capacity ratio, electrolyte residue, electrolyte composition data, and separator category of the battery sample to ensure the prediction accuracy of the random forest model.

[0121] Thus, in an embodiment of the present application, model training can be performed based on at least one of the cell thickness, cell width, cell height, cut-off voltage, positive electrode type, positive electrode plate composition data, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode plate composition data, negative electrode loading, negative electrode compaction, negative-to-positive electrode capacity ratio, electrolyte residual amount, electrolyte composition data, and diaphragm category of the battery sample, thereby ensuring the prediction accuracy of the model.

[0122] In certain embodiments of the present application, the performance test results include intermittent cycle test results of battery samples.

[0123] Specifically, in the embodiment of the present application, a random forest model suitable for the task of predicting battery intermittent cycle test results, that is, a test model, can be trained through the chemical and cell parameters of the battery samples and the intermittent cycle test results of the battery samples.

[0124] In one example, the intermittent cycle test result of the battery sample includes at least one of the energy retention rate, charge and discharge efficiency, and cycle life of the battery sample.

[0125] The energy retention rate can evaluate the capacity attenuation of the battery during intermittent cycling, and can be determined by comparing the discharge capacity of the battery at different cycle numbers with the initial discharge capacity.

[0126] The charge and discharge efficiency can evaluate the energy conversion efficiency of the battery during intermittent cycles and can be determined based on the actual amount of electricity charged and discharged during the charging and discharging process.

[0127] Cycle life evaluates the useful life of a battery, which can be determined by the number of cycles the battery can complete before meeting a certain performance indicator (such as the capacity retention rate drops to 80%).

[0128] Thus, in the embodiment of the present application, a random forest model suitable for the task of predicting battery intermittent cycle test results can be trained through the chemical and cell parameters of the battery samples and the intermittent cycle test results of the battery samples, that is, a test model, thereby avoiding the need to conduct real intermittent cycle tests on the battery to be tested to obtain the intermittent cycle test results of the battery to be tested, thereby reducing the proportion of intermittent cycle tests in the battery performance test link to a certain extent, and thereby reducing the time cost of the battery test link, thereby improving the battery test efficiency.

[0129] See also Figure 2 In certain embodiments of the present application, the sub-step of determining the training data based on the battery sample data in step 01 includes:

[0130] 010: Preprocess the battery sample data and determine the preprocessed battery sample data;

[0131] 011: Determine training data based on the preprocessed battery sample data.

[0132] Specifically, in an embodiment of the present application, the electronic device may perform preprocessing, such as denoising, on the battery sample data after obtaining the battery sample data, thereby determining the preprocessed battery sample data, and then determine the training data for model training based on the preprocessed battery sample data.

[0133] In one example, in an embodiment of the present application, preprocessing of the battery sample data may include noise reduction processing to reduce noise in the battery sample data. It is understood that since the random forest model is sensitive to noise in the data, the battery sample data may be preprocessed to reduce noise, thereby reducing the negative impact of noise in the battery sample data on the random forest model.

[0134] In this way, in the embodiment of the present application, the battery sample data can be preprocessed to determine the preprocessed battery sample data, and the training data can be determined based on the preprocessed battery sample data, so that the training data can be determined based on the preprocessed battery sample data, thereby ensuring the effectiveness and reliability of the training data to a certain extent, and further ensuring the performance of the random forest model trained based on the training data.

[0135] In certain embodiments of the present application, the preprocessing includes at least one of blank field filling, preset field conversion, and preset field deletion.

[0136] Specifically, in the embodiment of the present application, at least one of blank field filling, preset field conversion and preset field deletion can be performed to complete the preprocessing of the battery sample data, thereby reducing the noise in the battery sample data.

[0137] In one example, the performance test results of the battery samples include the intermittent cycle test results of the battery samples, and the test time is required to be 136 days. Then, among the intermittent cycle test results of all battery samples, the intermittent cycle test results that are less than 136 days are deleted, thereby achieving the deletion of the preset field.

[0138] In one example, the intermittent cycle test results of the battery sample include the energy retention rate of the battery sample, and it is set that the battery sample passes the intermittent cycle test when the energy retention rate of the battery sample is 70% or above, otherwise it fails the test if it is below 70%. Furthermore, the intermittent cycle test results of the battery sample with an energy retention rate of 70% or above can be set to 1, and the intermittent cycle test results of the battery sample with an energy retention rate of less than 70% can be set to 0, thereby realizing the preset field conversion.

[0139] In one example, for the batch, positive electrode, negative electrode, and separator categories of the battery samples, one-hot encoding processing can be performed on the batch, positive electrode, negative electrode, and separator categories of the battery samples, thereby achieving preset field conversion.

[0140] In one example, the battery sample data includes the chemical and cell parameters of the battery sample, and missing values ​​are filled in for the chemical and cell parameters. For example, the missing value of the SP (Super P) content in the electrode composition is filled with 0, and the missing value of the electrolyte additive content is filled with 0, thereby achieving blank field filling.

[0141] Thus, in the embodiment of the present application, at least one of blank field filling, preset field conversion, and preset field deletion may be performed on the battery sample data, thereby completing the preprocessing of the battery sample data.

[0142] In certain embodiments of the present application, determining the training data based on the pre-processed battery sample data includes:

[0143] Determine multiple groups of pre-processed battery sample data, and determine training data based on the multiple groups of pre-processed battery sample data.

[0144] The processor of the embodiment of the present application is further configured to determine multiple sets of pre-processed battery sample data, and determine training data based on the multiple sets of pre-processed battery sample data.

[0145] Specifically, in embodiments of the present application, after preprocessing the battery sample data to obtain preprocessed battery sample data, the preprocessed battery sample data can be divided into multiple groups. Furthermore, the electronic device of embodiments of the present application can determine training data for random forest model training based on the multiple groups of preprocessed battery sample data.

[0146] For example, when 10 groups of pre-processed battery sample data are obtained, the first group can be used as a test set, and the second to tenth groups can be used as training sets, thereby obtaining training data consisting of the test set and the training set.

[0147] Thus, in the embodiment of the present application, multiple sets of pre-processed battery sample data can be determined, and training data can be determined based on the multiple sets of pre-processed battery sample data, thereby achieving the determination of training data and ensuring the quality of training data to a certain extent.

[0148] In certain embodiments of the present application, the step of determining training data based on multiple sets of pre-processed battery sample data includes:

[0149] A portion of the preprocessed battery sample data is determined as a training set, and a portion of the preprocessed battery sample data is determined as a test set, wherein the preprocessed battery sample data included in the training set and the preprocessed battery sample data included in the test set are at least partially staggered, and the training data is determined based on the test set and the training set.

[0150] The processor of the embodiment of the present application is also used to determine a part of the group of preprocessed battery sample data as a training set, and determine a part of the group of preprocessed battery sample data as a test set, wherein the preprocessed battery sample data included in the training set and the preprocessed battery sample data included in the test set are at least partially staggered, and the training data is determined based on the test set and the training set.

[0151] Specifically, in an embodiment of the present application, when preprocessing multiple groups of battery sample data to obtain multiple groups of preprocessed battery sample data, the electronic device can use some groups of preprocessed battery sample data as training sets, and the remaining groups of preprocessed battery sample data as test sets, thereby completing the construction of training data.

[0152] For example, assuming that the preprocessed battery sample data includes N groups, these N groups can be divided into 10 parts, resulting in 10 N / 10 parts. Then, one of the N / 10 parts is used as the test set, and the remaining 9 N / 10 parts are used as the training set.

[0153] Furthermore, it can be understood that the preprocessed battery sample data in the test set and the preprocessed battery sample data in the training set can be at least partially staggered, thereby avoiding the use of the preprocessed battery sample data used when training the model in the performance testing process of the model. Furthermore, in the process of testing the model using the test set, the trained model can perform inference and prediction on the preprocessed battery sample data that has not been processed during training, so that the inference and prediction results of the model can more accurately reflect the performance of the model itself, thereby ensuring the accurate determination of the model performance.

[0154] Thus, in the embodiment of the present application, some groups of pre-processed battery sample data can be determined as training sets, and some groups of pre-processed battery sample data can be determined as test sets, thereby completing the construction of training data and ensuring the robustness of subsequent model training steps.

[0155] In certain embodiments of the present application, the steps of determining a portion of the pre-processed battery sample data as a training set and determining a portion of the pre-processed battery sample data as a test set include:

[0156] Multiple sets of pre-processed battery sample data are subjected to multiple preset processing to determine multiple sets of training data, wherein the preset processing includes: determining some sets of pre-processed battery sample data as training sets, and determining some sets of pre-processed battery sample data as test sets, the pre-processed battery sample data included in the training sets and the pre-processed battery sample data included in the test sets are at least partially staggered, and one set of training data does not completely overlap with another set of training data.

[0157] The processor of the embodiment of the present application is also used to perform multiple preset processing on multiple groups of preprocessed battery sample data to determine multiple training data, wherein the preset processing includes: determining some groups of preprocessed battery sample data as training sets, and determining some groups of preprocessed battery sample data as test sets, the preprocessed battery sample data included in the training set and the preprocessed battery sample data included in the test set are at least partially staggered, and one set of training data does not completely overlap with another set of training data.

[0158] Specifically, in the embodiment of the present application, the electronic device can perform multiple preset processes on multiple groups of pre-processed battery sample data, thereby obtaining multiple sets of training data, and thus obtaining training data used for model training.

[0159] In one example, a 10-fold cross-validation approach can be used to complete dataset construction to avoid model overfitting. Specifically, all N groups of preprocessed battery sample data are first divided into 10 non-overlapping subsets. Then, each subset includes N / 10 groups of preprocessed battery sample data. Let these ten subsets be X_train1, X_train2, ..., X_train10, respectively.

[0160] Furthermore, X_train1 can be used as a test set, and X_train2, ..., X_train10 can be used as training sets, thereby obtaining the first set of training data.

[0161] Then, X_train2 is used as the test set, and X_train1, X_train3, ..., X_train10 are used as the training set, thereby obtaining the second set of training data.

[0162] Then, X_train3 is used as the test set, and X_train1, X_train2, X_train4, ..., X_train10 are used as the training set, thus obtaining the third set of training data.

[0163] And so on, 10 sets of training data are obtained, and then these 10 sets of training data can be integrated into complete training data for model training.

[0164] It is understandable that the test sets for any two training data sets do not completely overlap. For example, the test set for the first training data set is X_train1, and the test set for the second training data set is X_train2. It is also understandable that the incomplete overlap between any two training data sets ensures that the model is trained differently based on different training data sets, thereby improving the model training effect to a certain extent.

[0165] Thus, in the embodiment of the present application, multiple sets of pre-processed battery sample data may be subjected to multiple preset processes to determine multiple sets of training data, thereby achieving the construction of training data.

[0166] See also Figure 3 In certain embodiments of the present application, the random forest model includes multiple, and then, step 01 includes:

[0167] 012: According to the training set, train the random forest model and determine the random forest model after training;

[0168] 013: Determine the model performance data of the trained random forest model based on the test set and the trained random forest model;

[0169] 014: Based on the model performance data, determine at least one of the trained random forest models as a testing model.

[0170] The processor in the embodiment of the present application is also used to train the random forest model based on the training set to determine the trained random forest model, and to determine the model performance data of the trained random forest model based on the test set and the trained random forest model, and to determine at least one of the trained random forest models as a test model based on the model performance data.

[0171] Specifically, in the embodiments of the present application, the electronic device may train a random forest model using a training set in the training data, and after the random forest model training is completed, the electronic device may perform a performance test on the trained random forest model using a test set in the training data. Finally, after completing the training and performance testing of each random forest model, the electronic device may determine at least one of the multiple trained random forest models as a test model for battery performance testing based on the performance test results of each random forest model.

[0172] For example, when there are 10 random forest models, the implementation method of the present application can train and test these 10 random forest models separately according to the training data, so as to obtain performance test results of 10 trained random forest models. Then, the electronic device can determine one or more of the 10 trained random forest models as test models for battery performance testing based on the performance test results of these 10 trained random forest models.

[0173] Thus, in an embodiment of the present application, the random forest model can be trained based on the training set to determine the trained random forest model, and the model performance data of the trained random forest model can be determined based on the test set and the trained random forest model, and at least one of the multiple trained random forest models can be determined as the test model based on the model performance data of each trained random forest model, thereby completing the training and testing of the test model, so that the test model can be robustly determined.

[0174] In certain embodiments of the present application, the model performance data includes recognition accuracy, and step 014 includes:

[0175] The trained random forest model with the highest recognition accuracy among the trained random forest models is determined as the test model.

[0176] The processor of the embodiment of the present application is also used to determine the trained random forest model with the highest recognition accuracy among the trained random forest models as the test model.

[0177] Specifically, in the embodiment of the present application, according to the recognition accuracy (Accuracy) of each trained random forest model, the trained random forest model with the highest recognition accuracy among all the trained random forest models can be determined as the test model.

[0178] It can be understood that the recognition accuracy of the trained random forest model can be used to indicate the accuracy of the trained random forest model in predicting the performance test results of the battery.

[0179] Specifically, in an embodiment of the present application, the battery sample data includes the chemical and cell parameters of the battery sample (let the chemical and cell parameters be X), and the performance test results of the battery sample (let the performance test results be Y). Furthermore, the recognition accuracy of the trained random forest model can be understood as the probability that the trained random forest model predicts the performance test result Y based on the chemical and cell parameters X of the battery sample.

[0180] Thus, in the embodiment of the present application, the trained random forest model with the highest recognition accuracy among multiple trained random forest models can be determined as the test model, thereby ensuring that the test model can effectively and reliably predict the battery performance test results.

[0181] See also Figure 4 In certain embodiments of the present application, the battery sample data includes battery parameters and performance test results. Then, step 013 includes:

[0182] 0130: Determine the performance prediction results of the trained random forest model for the battery sample based on the trained random forest model and the battery parameters of the battery sample;

[0183] 0131: Determine the model performance data based on the performance prediction results of the trained random forest model for each battery sample and the performance test results of each battery sample.

[0184] The processor of the embodiment of the present application is also used to determine the performance prediction results of the trained random forest model for the battery sample based on the trained random forest model and the battery parameters of the battery sample, and to determine the model performance data based on the performance prediction results of the trained random forest model for each battery sample and the performance test results of each battery sample.

[0185] Specifically, in an embodiment of the present application, the battery sample data includes the chemical and cell parameters of the battery sample (let the chemical and cell parameters be X), and the performance test results of the battery sample (let the performance test results be Y). Furthermore, the electronic device can predict the performance test results of the battery sample based on the chemical and cell parameters X of the battery sample according to the trained random forest model, thereby determining the performance prediction result of the battery sample (let the performance prediction result be Y'). Finally, the electronic device can determine the model performance data of the trained random forest model based on the performance test result Y and the performance prediction result Y' of the battery sample.

[0186] For example, when there are 50 battery samples, the electronic device can determine the prediction accuracy of the trained random forest model based on the performance prediction results Y' and performance test results Y of these 50 battery samples. For example, when the performance prediction results Y' and performance test results Y of 35 battery samples among these 50 battery samples are consistent, and the performance prediction results Y' and performance test results Y of the other 15 battery samples are inconsistent, then the prediction accuracy of the trained random forest model is 35 / 50, that is, 0.7.

[0187] Thus, in the embodiment of the present application, the performance prediction results of the trained random forest model for the battery sample can be determined based on the trained random forest model and the battery parameters of the battery sample, and the model performance data can be determined based on the performance prediction results of the trained random forest model for each battery sample and the performance test results of each battery sample, thereby achieving the determination of the model performance data, and then the test model can be robustly determined based on the model performance data.

[0188] See also Figure 5 In certain embodiments of the present application, the test model for battery performance testing further includes:

[0189] 02: Adjust the pre-set hyperparameters and determine at least one adjusted hyperparameter;

[0190] 03: Determine multiple random forest models based on at least one tuned hyperparameter.

[0191] The processor of the embodiment of the present application is also used to adjust the pre-set hyperparameters, determine at least one adjusted hyperparameter, and determine multiple random forest models based on the at least one adjusted hyperparameter.

[0192] Specifically, in an embodiment of the present application, one or more adjustments may be made based on pre-specified initial hyperparameters to determine one or more adjusted hyperparameters, and then multiple random forest models to be trained may be determined based on the one or more adjusted hyperparameters.

[0193] Thus, in an embodiment of the present application, the pre-set hyperparameters can be adjusted to determine at least one adjusted hyperparameter, and multiple random forest models can be determined based on the at least one adjusted hyperparameter, thereby achieving the determination of multiple random forest models.

[0194] In certain embodiments of the present application, the hyperparameters include the number of decision trees and / or the depth of the decision trees.

[0195] Specifically, in the embodiment of the present application, the number of decision trees and / or the depth of decision trees can be preset, and the electronic device can construct multiple random forest models to be trained based on the preset number of decision trees and / or the depth of decision trees.

[0196] In one example, n_estimators (ie, the number of decision trees) can be set to 25, max_depth (decision tree depth) can be set to 6, and random (random value) can be set to 80, so that the electronic device generates multiple random forest models to be trained based on this.

[0197] Thus, in the embodiment of the present application, the number of decision trees and / or the depth of decision trees can be preset to construct multiple random forest models to be trained.

[0198] To more clearly illustrate the implementation of this application, please refer to Figure 6 、 Figure 7 and Figure 8 , Figure 6 、 Figure 7 and Figure 8 These are schematic diagrams of application scenarios in certain embodiments of the present application. Figure 6 As shown, in one example, a first acquisition unit, a data preprocessing unit, a training unit, a second acquisition unit and a prediction unit are provided in the electronic device (or electronic device). Among them, the first acquisition unit is used to obtain the historical chemical system and cell design parameters of the battery. The data preprocessing unit is used to preprocess the battery parameters and generate a training set based on the preprocessed battery parameters. The training unit is used to train the random forest classification model using the training set to obtain a prediction model of the relationship between the battery parameters and whether the battery passes the intermittent cycle test. The second acquisition unit is used to obtain the chemical system and cell design parameters of the battery to be tested, and use them as a prediction set for predicting whether the battery passes the test. The prediction unit is used to generate a prediction result of the battery intermittent cycle test based on the prediction set and the prediction model.

[0199] In one example, the preprocessing unit further includes a data processing unit, a statistics unit, and a selection unit. The data processing unit is configured to perform data cleaning, denoising, and feature extraction on battery parameters to generate training data. The statistics unit is configured to annotate the training data and establish a correspondence between battery parameters and whether the battery passes or fails the test. The selection unit is configured to randomly select a preset amount of training data to generate a training set.

[0200] In one example, the electronic device (or electronic device) is further provided with an optimization unit. The optimization unit is configured to generate a test set based on the preprocessed battery parameters, input the test set into a prediction model to obtain a prediction result, compare the prediction result with the actual battery result to obtain an error index, and optimize the prediction performance of the prediction model based on the error index.

[0201] As well as Figure 7As shown, in an embodiment of the present application, the design data of the battery sample and the test data of the intermittent cycle can be collected in advance. Among them, the design data includes battery cell thickness, battery cell width, battery cell height, batch, voltage, positive electrode type, positive electrode sheet composition information, positive electrode load, positive electrode compaction, negative electrode type, negative electrode sheet composition information, negative electrode load, negative electrode compaction, N / P ratio, electrolyte composition, diaphragm type, etc. The intermittent cycle test data is the 136-day energy retention rate of the battery sample. In order to facilitate the explanation of each battery sample, the design information of each battery sample is merged with the energy retention rate data of the intermittent cycle test for 136 days.

[0202] Next, the design information was combined with the energy retention data from 136 days of intermittent cycling tests for data cleaning. For example, for intermittent cycling test data, data with less than 136 days of testing was deleted. In another example, missing values ​​in the battery design data were filled with 0, such as the missing values ​​for the SP content in the electrode composition and the missing values ​​for the electrolyte additive content. One-hot encoding was used for batch, positive electrode, negative electrode, and separator data.

[0203] Next, feature data is selected for the design information of the battery sample. In one example, all design information can be used as feature data, or in other words, the feature data is the battery design information, including cell thickness, cell width, cell height, batch, voltage, positive electrode type, positive electrode sheet composition information, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode sheet composition information, negative electrode loading, negative electrode compaction, N / P ratio, residual amount, electrolyte composition, and separator type.

[0204] Subsequently, the data is divided into a training set X_train and a test set X_test with a ratio of 0.7:0.3.

[0205] Then, the random forest model is trained and tested based on the training set X_train and the test set X_test.

[0206] The training and testing process uses a 10-fold cross-validation approach, which involves: 1) Divide the entire training set X_train into 10 non-overlapping subsets. Assuming the number of training examples in X_train is n, each subset has n / 10 training examples, and the corresponding subsets are {X_train1, X_train2, …, X_train10}. 2) Then, select the first subset as the test set, and the other 9 as the training set. 3) Then, train the random forest model based on the training and test sets, and calculate the classification accuracy. 4) Then, repeat steps 2) and 3) 10 times, selecting a different subset as the test set each time. Next, calculate the average of the 10 prediction accuracy scores as the model's prediction accuracy. It is understood that the training set can be used to train the instantiated model, implemented using rfc.fit(X_train, y_train). It is also understandable that the test set data X_test can be used as the model input, and the trained model can be used for prediction to obtain the predicted value y_pred. Then, the prediction accuracy can be calculated based on the predicted value and the true value, such as through rfc.score(X_test,y_test).

[0207] After the training and testing process is complete, plot the learning curve for the number of tree models n_estimators and find the value of n_estimators that gives the highest accuracy. Also, use a grid search to find the value of the maximum depth max_depth.

[0208] Finally, based on the n_estimators corresponding to the highest accuracy and the maximum depth max_depth found, the final random forest model is determined, which is the above test model.

[0209] In order to clearly illustrate the application effect of the random forest model (or test model) in the embodiment of the present application on the battery performance test, the embodiment of the present application also provides a battery performance test solution based on the support vector machine classification model, namely:

[0210] Specifically, unlike training a random forest model, support vector machine training can include a data standardization step to address dimensional differences between data, thereby improving model training efficiency and performance. Specifically, columns other than the ID, Batch, Positive Electrode, Negative Electrode, Separator, and Retention Rate columns are selected and standardized (Standard). In one example, standardization is implemented using the StandardScaler().fit_transform function.

[0211] Similar to random forest model training, the preprocessed training data is divided into a training set X_train and a test set X_test with a ratio of 0.7:0.3. Furthermore, the kernel functions used are linear kernel, polynomial kernel, Gaussian radial kernel, and hyperbolic tangent kernel.

[0212] During support vector machine training, a 10-fold cross-validation approach is used to train the dataset. Specifically: 1. Divide the entire training set X_train into 10 disjoint subsets. Assuming the number of training examples in X_train is n, then each subset has n / 10 training examples, and the corresponding subsets are {X_train1, X_train2, …, X_train10}. 2. Select the first subset as the test set, and the other nine as the training set. 3. Train the classification model and calculate the classification accuracy using the cross_val_score function in sklearn. 4. Repeat steps 2 and 3 10 times, selecting a different subset as the test set each time. 5. Calculate the average of the 10 classification accuracy scores as the model's accuracy.

[0213] It's understandable that during training, the training set can be fed into the instantiated model for training. During testing, the test set data X_test can be fed into the trained model for prediction, yielding the predicted value y_pred. The predicted value y_pred is then compared with the true value to calculate its accuracy.

[0214] It's also understandable that both the trained random forest model and support vector machine model were based on random forest cross-validation. The random forest model achieved a prediction accuracy of 83%, and 87.5% in the test set. Table 2 shows a comparison with the support vector machine model, showing significant differences in accuracy between different kernel functions. The optimized Gaussian radial kernel achieved higher accuracy than the other kernel functions. The random forest classification model achieved significantly higher accuracy than the support vector machine model.

[0215] Table 2

[0216] Kernel Function Accuracy Accuracy after optimization Linear kernel 63％ 72.1％ Polynomial kernel 70.2％ - Gaussian radial kernel 70.2％ 76.6％

[0217] In addition, the embodiment of the present application also conducts a correlation test between the prediction accuracy and the size of the data set, such as Figure 8 As shown in the figure, as the amount of data increases, the prediction accuracy continues to improve, which shows that the amount of data has an important impact on the accuracy.

[0218] See also Figure 9 Corresponding to the above-mentioned test model for battery performance testing, the embodiment of the present application further provides a battery performance testing method, including:

[0219] 07: According to the battery data to be tested and the above-mentioned test model, determine the performance test result of the battery to be tested corresponding to the battery data to be tested.

[0220] The present application also provides an electronic device, comprising a memory and a processor. The battery performance testing method of the present application can be implemented by the electronic device of the present application. Specifically, the memory stores a computer program, and the processor is configured to determine a battery performance test result corresponding to the battery data and the test model described above based on the battery data.

[0221] Specifically, in an embodiment of the present application, the electronic device (or electronic device) can call a pre-trained test model when obtaining the data of the battery to be tested, that is, the battery data to be tested, so that the test model can predict the battery performance test results corresponding to the battery data to be tested based on the battery data to be tested. Furthermore, the battery performance test of the battery to be tested can be obtained without performing an actual battery performance test on the battery to be tested.

[0222] It is understandable that the training method of the test model can be determined based on the test model for battery performance testing described above, and to avoid repetition, it will not be repeated here.

[0223] In one example, the battery data to be tested includes chemical and cell parameters of the battery to be tested.

[0224] In one example, the chemical and cell parameters of the battery to be tested include cell thickness, cell width, cell height, cut-off voltage, positive electrode type, positive electrode plate composition data, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode plate composition data, negative electrode loading, negative electrode compaction, negative-to-positive electrode capacity ratio, electrolyte residual amount, electrolyte composition data and at least one of the diaphragm category.

[0225] In one example, the performance test results of the battery to be tested include intermittent cycle test results predicted by the test model based on the data of the battery to be tested.

[0226] Thus, in the embodiment of the present application, a pre-set random forest model can be trained to determine a test model suitable for battery performance testing. Then, the performance test results of the battery to be tested can be determined based on the test model determined by the trained random forest model, thereby avoiding the need to conduct a real battery performance test on the battery to be tested to obtain the performance test results, such as intermittent cycle testing with a long test cycle, etc., thereby shortening the proportion of the battery performance test link in the entire battery development process to a certain extent, reducing battery development costs such as time costs, and improving battery development efficiency. In addition, because the test model is obtained by training based on the random forest model, the data required for model training is less than that for neural network models such as convolutional neural network models, the training difficulty is lower and the training cost is lower, making it more suitable for battery performance testing scenarios with higher real-time requirements.

[0227] An embodiment of the present application further provides an electronic device, which includes the above-mentioned electronic device.

[0228] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, it implements the above-mentioned test model for battery performance testing, or implements the above-mentioned battery performance testing method.

[0229] The present application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, is used as a test model for battery performance testing, or implements the above-mentioned battery performance testing method.

[0230] In the description of this specification, the descriptions with reference to the terms "particularly", "further", "particularly", "understandably", etc. are intended to mean that the specific features, structures, materials or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic expressions of the above terms are not intended to refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0231] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0232] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A test model for battery performance testing, characterized in that: include: The preset random forest model is trained based on the training data determined by the battery sample data to determine the test model.

2. The test model according to claim 1, characterized in that The battery sample data includes at least one of chemical and cell parameters of the battery sample and performance test results of the battery sample.

3. The test model according to claim 2, characterized in that The chemical and cell parameters include cell thickness, cell width, cell height, cut-off voltage, positive electrode type, positive electrode plate composition data, positive electrode loading, positive electrode compaction, negative electrode type, negative electrode plate composition data, negative electrode loading, negative electrode compaction, negative-to-positive electrode capacity ratio, electrolyte residual amount, electrolyte composition data and at least one of the diaphragm type.

4. The test model according to claim 3, characterized in that The performance test results include intermittent cycle test results of the battery samples.

5. The test model according to claim 1, characterized in that The training data determined based on the battery sample data includes: Performing preprocessing on the battery sample data to determine preprocessed battery sample data; The training data is determined according to the preprocessed battery sample data.

6. The test model according to claim 5, characterized in that The preprocessing includes at least one of blank field filling, preset field conversion and preset field deletion.

7. The test model according to claim 5, characterized in that The step of determining the training data based on the pre-processed battery sample data includes: Determine a plurality of groups of pre-processed battery sample data, and determine the training data according to the plurality of groups of pre-processed battery sample data.

8. The test model according to claim 7, characterized in that: The determining the training data according to the plurality of groups of pre-processed battery sample data includes: Determine a portion of the preprocessed battery sample data as a training set, and determine a portion of the preprocessed battery sample data as a test set, wherein the preprocessed battery sample data included in the training set and the preprocessed battery sample data included in the test set are at least partially staggered, and determine the training data based on the test set and the training set.

9. The test model according to claim 8, characterized in that The step of determining a portion of the pre-processed battery sample data as a training set and determining a portion of the pre-processed battery sample data as a test set includes: Perform multiple preset processing on multiple groups of the preprocessed battery sample data to determine multiple sets of the training data, wherein the preset processing includes: determining some groups of the preprocessed battery sample data as training sets, and determining some groups of the preprocessed battery sample data as test sets, the preprocessed battery sample data included in the training sets and the preprocessed battery sample data included in the test sets are at least partially staggered, and one set of the training data does not completely overlap with another set of the training data.

10. The test model according to claim 1, characterized in that: The random forest model includes multiple models, the training data includes a training set and a test set, and the training data determined according to the battery sample data is used to train the preset random forest model to determine the test model, including: Training the random forest model according to the training set to determine a trained random forest model; Determining model performance data of the trained random forest model based on the test set and the trained random forest model; At least one of the trained random forest models is determined as the test model based on the model performance data.

11. The test model according to claim 10, characterized in that: The model performance data includes recognition accuracy, and determining at least one of the trained random forest models as the test model based on the model performance data includes: The trained random forest model with the highest recognition accuracy among the trained random forest models is determined as the test model.

12. The test model according to claim 10, characterized in that The battery sample data includes battery parameters and performance test results, and determining the model performance data of the trained random forest model based on the test set and the trained random forest model includes: Determining a performance prediction result of the trained random forest model for the battery sample based on the trained random forest model and the battery parameters of the battery sample; The model performance data is determined based on the performance prediction results of the trained random forest model for each of the battery samples and the performance test results of each of the battery samples.

13. The test model according to claim 1, characterized in that Adjusting a preset hyperparameter to determine at least one adjusted hyperparameter; The random forest model is determined based on at least one of the adjusted hyperparameters.

14. The test model according to claim 13, characterized in that The hyperparameters include the number of decision trees and / or the depth of the decision trees.

15. A battery performance testing method, characterized in that: include: According to the battery data to be tested and the test model according to any one of claims 1 to 14, a performance test result of the battery to be tested corresponding to the battery data to be tested is determined.

16. An electronic device, characterized in that: The invention comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the test model described in any one of claims 1 to 14 is implemented, and / or the battery performance test method described in claim 15 is implemented.

17. An electronic device, characterized in that: The electronic device comprises the apparatus according to claim 16.

18. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by one or more processors, it implements the test model described in any one of claims 1 to 14, and / or implements the battery performance testing method described in claim 15.

19. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, it implements the test model described in any one of claims 1 to 14, and / or implements the battery performance testing method described in claim 15.

Images