Battery capacity detection method, control device, electronic device and storage medium

Abstract

The invention discloses a battery capacity detection method, a control device, an electronic device and a computer readable storage medium. The method comprises the steps of obtaining target production line data of a to-be-detected battery, determining a capacity category prediction result and a capacity value prediction result of the to-be-detected battery according to the target production line data of the to-be-detected battery, and a pre-constructed capacity classification model and a pre-constructed capacity value prediction model, and determining the capacity of the to-be-detected battery according to the capacity category prediction result and the capacity value prediction result. And determining the capacity of the to-be-detected battery to complete capacity detection. Thus, the battery capacity detection can be determined through the target production line data related to at least one of the active lithium loss characteristic, the internal resistance difference characteristic and the capacity exerting characteristic of the battery to be detected, the accuracy of the capacity detection result is guaranteed, the battery capacity can be evaluated without discharging the battery, the capacity detection efficiency is improved, and the battery capacity detection efficiency is improved. And the energy consumption in the capacity detection link is reduced.

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a battery capacity detection method, a control device, an electronic device and a computer-readable storage medium. Background Art

[0002] Lithium batteries are widely used in electric vehicles and other fields due to their advantages such as high energy density and low cost. Battery capacity is one of the key indicators to measure the performance of lithium batteries, which is related to the mileage and life of lithium batteries. Therefore, it is necessary to accurately detect whether the capacity of lithium batteries is qualified. However, in related technologies, the capacity detection process of lithium batteries usually uses a capacity cabinet to fully charge and discharge the lithium battery, which has low detection efficiency and high energy consumption. Summary of the invention

[0003] The present application provides a battery capacity detection method, a control device, an electronic device and a computer-readable storage medium.

[0004] The present application provides a method for detecting the capacity of a battery, comprising:

[0005] Acquire target production line data of the battery to be tested, wherein the target production line data is related to at least one of an active lithium loss characteristic, an internal resistance difference characteristic, and a capacity utilization characteristic of the battery to be tested;

[0006] Determine a capacity category prediction result and a capacity value prediction result of the battery to be tested according to target production line data of the battery to be tested, and a pre-built capacity classification model and a capacity value prediction model;

[0007] The capacity of the battery to be tested is determined according to the capacity category prediction result and the capacity value prediction result, and the capacity detection is completed.

[0008] Thus, in the implementation mode of the present application, the capacity detection of the battery to be tested can be determined by the target production line data related to at least one of the active lithium loss characteristics, internal resistance difference characteristics and capacity utilization characteristics of the battery to be tested, so that the capacity detection of the battery to be tested can be supported by reliable data, the accuracy of the capacity detection result can be guaranteed, and the traditional capacity detection scheme implemented by the capacity cabinet and full filling and discharging can be abolished. The battery capacity can be evaluated without discharging the battery, the capacity detection efficiency can be improved, and the energy consumption in the capacity detection link can be reduced. In addition, the capacity detection result of the battery to be tested can be determined by the capacity category prediction result and the capacity value prediction result output by the two models of the pre-constructed capacity classification model and the capacity value prediction model, respectively, thereby ensuring the improvement of the prediction detection of the battery capacity.

[0009] In certain embodiments of the present application, the capacity category prediction result includes qualified and unqualified, and the capacity category prediction result and the capacity value prediction result of the battery to be tested are determined according to the target production line data of the battery to be tested, and the pre-built capacity classification model and capacity value prediction model, including:

[0010] Determining the capacity category prediction result according to the target production line data and the capacity classification model;

[0011] When the capacity category prediction result is qualified, the capacity value prediction result is determined according to the target production line data of the battery to be tested and the capacity value prediction model.

[0012] Thus, in the implementation mode of the present application, when the capacity classification model outputs a qualified capacity category prediction result based on the target production line data, the capacity value prediction model can be called to determine the capacity value prediction result of the battery to be tested, thereby reducing the resource consumption of the model call and avoiding invalid resource consumption to a certain extent.

[0013] In certain embodiments of the present application, the capacity category prediction result includes qualified and unqualified, and the capacity of the battery to be tested is determined according to the capacity category prediction result and the capacity value prediction result, and the capacity detection is completed, including:

[0014] When the capacity category prediction result is qualified and the capacity value prediction result is greater than or equal to a preset capacity threshold, determining that the capacity of the battery to be tested is qualified; and / or

[0015] When the capacity category prediction result is unqualified, or the capacity value prediction result is less than the preset capacity threshold, it is determined that the capacity of the battery to be tested is unqualified.

[0016] Thus, in the embodiments of the present application, the capacity of the battery to be tested can be determined to be qualified when the capacity category prediction result is qualified and the capacity value prediction result is greater than or equal to the preset capacity threshold, and / or the capacity of the battery to be tested can be determined to be unqualified when the capacity category prediction result is unqualified or the capacity value prediction result is less than the preset capacity threshold, thereby ensuring the reliability and accuracy of the battery capacity detection results to a certain extent.

[0017] In certain embodiments of the present application, the method further comprises:

[0018] Obtaining a capacity label and a plurality of first production line data of a first battery sample;

[0019] Processing the capacity tag and the first production line data to determine a correlation analysis result between the capacity tag and the first production line data;

[0020] The target production line data among the plurality of first production line data are determined according to the correlation analysis processing result of each type of the first production line data.

[0021] Thus, in the embodiment of the present application, the capacity label and multiple first production line data of the first battery sample can be obtained. Then, the capacity label and the first production line data are processed to determine the correlation analysis result between the capacity label and the first production line data, and according to the correlation analysis processing result of each first production line data, the target production line data among the multiple first production line data is determined, so that the correlation between the target production line data and the battery capacity is guaranteed.

[0022] In certain embodiments of the present application, the step of obtaining the capacity label and a plurality of first production line data of the first battery sample includes:

[0023] Acquire the capacity label and a plurality of second production line data of the first battery sample;

[0024] The second production line data is cleaned to determine the first production line data, wherein the data cleaning includes data length alignment and / or outlier elimination.

[0025] Thus, in the embodiment of the present application, a data cleaning operation may be performed on the acquired second production line data to obtain the first production line data, thereby ensuring the validity and reliability of the first production line data and the target production line data.

[0026] In certain embodiments of the present application, the correlation analysis result includes a first sub-result, and the processing of the capacity label and the first production line data to determine the correlation analysis result of the capacity label and the first production line data includes:

[0027] The capacity label and the first production line data are subjected to scatter matrix analysis to determine the first sub-result.

[0028] Thus, in the embodiment of the present application, the correlation between the capacity label and each first production line data can be analyzed through a matrix scatter plot.

[0029] In certain embodiments of the present application, the correlation analysis result includes a second sub-result, and the processing of the capacity label and the first production line data to determine the correlation analysis result of the capacity label and the first production line data includes:

[0030] The assumption that there is a correlation between the capacity label and the first production line data is verified to obtain the second sub-result.

[0031] Thus, in the embodiment of the present application, the assumption that there is a correlation between the capacity tag and the first production line data can be verified, thereby analyzing the correlation between each capacity tag and each first production line data.

[0032] In certain embodiments of the present application, the correlation analysis result includes a third sub-result, and the processing of the capacity label and the first production line data to determine the correlation analysis result of the capacity label and the first production line data includes:

[0033] The correlation coefficient between the capacity label and the first production line data is calculated to obtain the third sub-result.

[0034] Thus, in the embodiment of the present application, the correlation coefficient between the capacity tag and the first production line data may be calculated to analyze the correlation between the capacity tag and the first production line data.

[0035] In certain embodiments of the present application, the correlation analysis result includes a fourth sub-result, and the processing of the capacity label and the first production line data to determine the correlation analysis result of the capacity label and the first production line data includes:

[0036] Main effect analysis is performed on the capacity label and the first production line data to obtain the fourth sub-result.

[0037] Thus, in the implementation mode of the present application, the significance of the impact of each first production line data on the capacity can be determined by the main effect analysis method, thereby determining the correlation between the capacity label and each first production line data.

[0038] In certain embodiments of the present application, the correlation analysis result includes a plurality of sub-results, and determining the target production line data among the plurality of first production line data according to the correlation analysis processing result of each first production line data includes:

[0039] Among the plurality of first production line data, first production line data in which each of the sub-results satisfies a corresponding preset condition is determined as the target production line data.

[0040] Thus, in the embodiment of the present application, among a plurality of first production line data, the first production line data in which each sub-result satisfies the corresponding preset conditions can be determined as the target production line data, thereby ensuring the reliability and validity of the target production line data.

[0041] In certain embodiments of the present application, the step of training the capacity classification model includes:

[0042] Acquire a first data set, wherein the first data set includes target production line data and capacity category labels of a second battery sample;

[0043] Model training is performed according to the first data set to determine the capacity classification model.

[0044] Thus, in an embodiment of the present application, a first data set may be obtained and model training may be performed based on the first data set to determine a capacity classification model, so as to obtain a capacity classification model that can output capacity category prediction results based on the input of target production line data through supervised training.

[0045] In certain embodiments of the present application, performing model training according to the first data set to determine the capacity classification model includes:

[0046] Determine a first training set and a first test set according to the first data set;

[0047] Training a first model according to the first training set;

[0048] Determining the classification accuracy of the first model after training according to the first test set;

[0049] When the classification accuracy is greater than or equal to a first threshold, the trained first model is determined as the capacity classification model.

[0050] Thus, in an embodiment of the present application, a first training set and a first test set can be determined based on a first data set, and a first model can be trained based on the first training set, and a classification accuracy of the trained first model can be determined based on the first test set, and when the classification accuracy is greater than or equal to a first threshold, the trained first model is determined as a capacity classification model, thereby achieving training of the first model.

[0051] In certain embodiments of the present application, the method further comprises:

[0052] Adjust the hyperparameters of the first model, wherein the hyperparameters include at least one of the number of tree models, learning rate, maximum depth of decision tree, regularization term weight, random seed, random sampling ratio of samples, and random sampling ratio of features.

[0053] Thus, in the implementation manner of the present application, the hyperparameters of the first model may be adjusted before the training of the first model begins, so as to improve the training effect of the first model to a certain extent.

[0054] In certain embodiments of the present application, the training step of the capacity value prediction model includes:

[0055] Acquire a second data set, wherein the second data set includes target production line data and capacity value labels of a third battery sample;

[0056] Model training is performed according to the second data set to determine the capacity value prediction model.

[0057] Thus, in an embodiment of the present application, a first data set may be obtained and model training may be performed based on the first data set to determine a capacity classification model, so as to obtain a capacity value prediction model that can output a capacity value prediction result based on the input of target production line data through supervised training.

[0058] In certain embodiments of the present application, performing model training according to the second data set to determine the capacity value prediction model includes:

[0059] Determine a second training set and a second test set according to the second data set;

[0060] Training a second model based on the second training set;

[0061] Determining a prediction error parameter of the trained second model according to the second test set;

[0062] When the prediction error parameter is less than or equal to a second threshold, the trained second model is determined as the capacity value prediction model.

[0063] Thus, in an embodiment of the present application, a second training set and a second test set can be determined based on the second data set, and a second model can be trained based on the second training set, and a prediction error parameter of the trained second model can be determined based on the second test set, and when the prediction error parameter is less than or equal to a second threshold, the trained second model is determined as a capacity value prediction model, thereby realizing the training of the second model.

[0064] In certain embodiments of the present application, the method further comprises:

[0065] Adjust the hyperparameters of the second model, wherein the hyperparameters include at least one of the number of tree models, learning rate, maximum depth of decision tree, regularization term weight, random seed, random sampling ratio of samples, and random sampling ratio of features.

[0066] Thus, in the implementation manner of the present application, the hyperparameters of the second model may be adjusted before the training of the second model begins, so as to improve the training effect of the second model to a certain extent.

[0067] In certain embodiments of the present application, the target production line data includes a plurality of production line parameters, and the method further includes:

[0068] According to the battery capacity prediction model, the contribution of each of the production line parameters to the battery capacity prediction model is determined to determine the correlation between each of the production line parameters and the battery capacity, wherein the battery capacity prediction model is pre-trained and can predict the capacity value of the battery to be tested based on the target production line data of the battery to be tested.

[0069] Thus, in the embodiment of the present application, the contribution of each production line parameter to the battery capacity prediction model can be determined based on the battery capacity prediction model to explain the contribution of each production line parameter to the capacity classification model and the capacity value prediction model.

[0070] The embodiment of the present application provides a control device, including a transceiver unit and a processing unit;

[0071] The transceiver unit is configured to obtain target production line data of the battery to be tested, wherein the target production line data is related to the capacity performance characteristics of the battery to be tested;

[0072] The processing unit is configured to determine the capacity category prediction result and the capacity value prediction result of the battery to be tested according to the target production line data of the battery to be tested, and a pre-built capacity classification model and a capacity value prediction model, and determine the capacity of the battery to be tested according to the capacity category prediction result and the capacity value prediction result to complete the capacity detection.

[0073] An embodiment of the present application provides an electronic device, including the above-mentioned control device.

[0074] An embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed by one or more processors, the above-mentioned battery capacity detection method is implemented.

[0075] The control device, electronic device and computer-readable storage medium provided in the embodiments of the present application enable the capacity detection of the battery to be tested to be determined by target production line data related to at least one of the active lithium loss characteristics, internal resistance difference characteristics and capacity utilization characteristics of the battery to be tested, so that the capacity detection of the battery to be tested can be supported by reliable data, the accuracy of the capacity detection result can be guaranteed, and the traditional capacity detection scheme implemented by the capacity cabinet and full filling and discharging can be abolished. The battery capacity can be evaluated without discharging the battery, the capacity detection efficiency can be improved, and the energy consumption in the capacity detection link can be reduced. In addition, the capacity detection result of the battery to be tested can be determined by the capacity category prediction result and the capacity value prediction result output by the two models of the pre-constructed capacity classification model and the capacity value prediction model, respectively, thereby ensuring the improvement of the prediction detection of the battery capacity.

[0076] 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 apparent from the description below, or will be learned through the practice of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] 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:

[0078] Figure 1 A schematic diagram of a flow chart of battery capacity detection in certain embodiments of the present application;

[0079] Figure 2 A schematic diagram of application scenarios in certain embodiments of the present application;

[0080] Figure 3 A schematic diagram of a flow chart of battery capacity detection in certain embodiments of the present application;

[0081] Figure 4 A schematic diagram of a flow chart of battery capacity detection in certain embodiments of the present application;

[0082] Figure 5 A schematic diagram of a flow chart of battery capacity detection in certain embodiments of the present application;

[0083] Figure 6 A schematic diagram of a flow chart of battery capacity detection in certain embodiments of the present application;

[0084] Figure 7 A schematic diagram of a flow chart of battery capacity detection in certain embodiments of the present application;

[0085] Figure 8 A schematic diagram of a flow chart of battery capacity detection in certain embodiments of the present application;

[0086] Fig. 9 A schematic diagram of a flow chart of battery capacity detection in certain embodiments of the present application;

[0087] Fig.10 This is a schematic diagram of the process of battery capacity detection in certain embodiments of the present application. DETAILED DESCRIPTION

[0088] The embodiments of the present application are described in detail below, and 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 from beginning to end. 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 cannot be understood as limiting the embodiments of the present application.

[0089] Lithium batteries have the advantages of high energy density and low cost, so they are widely used in electric vehicles and other fields. Furthermore, battery capacity is one of the key indicators to measure the performance of lithium batteries, which is related to the battery life and battery life. Therefore, in the lithium battery production process, the capacity of each battery cell must be accurately tested to eliminate the battery cells with unqualified capacity, thereby ensuring the performance indicators such as the battery life and service life of the lithium battery shipped out of the factory.

[0090] Generally, the capacity detection process in the lithium battery industry is mostly based on the capacity cabinet, that is, the lithium battery to be tested is fully charged and discharged through the capacity cabinet to achieve the capacity detection of the lithium battery to be tested. It is understandable that this detection process has the disadvantages of low detection efficiency, high energy consumption, equipment occupation and inventory cost. Therefore, the design of a new and reliable capacity detection method is an urgent need in the current lithium battery industry.

[0091] In a related technology, the lithium battery to be tested is charged and discharged, and then charged to the SOC (State of Charge) required for self-discharge screening, and then the discharge capacity corresponding to the formation discharge cut-off voltage and the discharge temperature at the end of discharge are obtained. Then, based on the temperature compensation formula, the discharge capacity is compensated by the discharge temperature calibration to obtain the compensated discharge capacity, and finally the compensated discharge capacity is brought into the pre-fitted capacity prediction formula to obtain the full discharge capacity of the lithium battery to be tested. However, it is understandable that this method still requires the lithium battery to be discharged, and cannot meet the urgent need to abolish the sub-capacity cabinet on the production line.

[0092] In another related technology, a formation charging sample data set and a fractional discharge capacity sample data set of a lithium battery are obtained, and then a data set to be trained is extracted from the formation charging sample data set. Then, at least one derivative feature in the data set to be trained is input into at least two initial training models for training to obtain at least two primary prediction models. Then, a secondary prediction model is obtained by integrated training based on at least two primary prediction models. Finally, after the formation charging data of the lithium battery to be predicted is input into the secondary prediction model, the predicted discharge capacity output of the lithium battery to be predicted is obtained. However, it can be understood that the lithium battery is a nonlinear and complex electrochemical system, and the detection process has a significant impact on the fractional discharge capacity, and this related technology does not fully consider the factors affecting the capacity, and the model is not interpretable, so it is impossible to accurately predict the capacity.

[0093] Based on the above problems that may be encountered, please refer to Figure 1 , the embodiment of the present application provides a battery capacity detection method, comprising:

[0094] 01: Obtain target production line data of the battery to be tested, wherein the target production line data is related to at least one of an active lithium loss characteristic, an internal resistance difference characteristic, and a capacity utilization characteristic of the battery to be tested;

[0095] 02: Determine the capacity category prediction result and capacity value prediction result of the battery to be tested based on the target production line data of the battery to be tested and the pre-built capacity classification model and capacity value prediction model;

[0096] 03: According to the capacity category prediction results and the capacity value prediction results, determine the capacity of the battery to be tested and complete the capacity test.

[0097] The embodiment of the present application provides a control device. The battery detection method of the embodiment of the present application can be implemented by the control device of the embodiment of the present application. Specifically, the control device includes a transceiver unit and a processing unit. Among them, the transceiver unit is configured to obtain the target production line data of the battery to be tested, and the target production line data is related to at least one of the active lithium loss characteristics, internal resistance difference characteristics and capacity utilization characteristics of the battery to be tested. The processing unit is configured to determine the capacity category prediction result and the capacity value prediction result of the battery to be tested according to the target production line data of the battery to be tested, and the pre-built capacity classification model and capacity value prediction model, and determine the capacity of the battery to be tested according to the capacity category prediction result and the capacity value prediction result, and complete the capacity detection.

[0098] Specifically, in an embodiment of the present application, an electronic device for detecting battery capacity (or a control device within the electronic device) can obtain target production line data of the battery to be tested, that is, production line data related to at least one of the active lithium loss characteristics, internal resistance difference characteristics, and capacity utilization characteristics of the battery to be tested in all production line data of the target production line data. Then, the target production line data is input into a pre-built and trained capacity classification model and a capacity value prediction model, thereby obtaining the capacity category prediction result and the capacity value prediction result of the battery to be tested. Finally, based on the capacity category prediction result and the capacity value prediction result of the battery to be tested, it is determined whether the capacity of the battery to be tested meets the expected production line process standard, thereby completing the capacity detection of the battery to be tested.

[0099] Thus, in the implementation mode of the present application, the capacity detection of the battery to be tested can be determined by the target production line data related to at least one of the active lithium loss characteristics, internal resistance difference characteristics and capacity utilization characteristics of the battery to be tested, so that the capacity detection of the battery to be tested can be supported by reliable data, the accuracy of the capacity detection result can be guaranteed, and the traditional capacity detection scheme implemented by the capacity cabinet and full filling and discharging can be abolished. The battery capacity can be evaluated without discharging the battery, the capacity detection efficiency can be improved, and the energy consumption in the capacity detection link can be reduced. In addition, the capacity detection result of the battery to be tested can be determined by the capacity category prediction result and the capacity value prediction result output by the two models of the pre-constructed capacity classification model and the capacity value prediction model, respectively, thereby ensuring the improvement of the prediction detection of the battery capacity.

[0100] To more clearly illustrate the implementation of this application, please refer to Figure 2 , Figure 2 These are all schematic diagrams of application scenarios in certain embodiments of the present application. Figure 2 The normal distribution used for capacity detection error has frequency on the ordinate and error on the scale. The mean is 0.000007714, the standard deviation is 0.0001762, and the total number N is 250000.

[0101] Furthermore, if Figure 2 As shown, in the embodiment of the present application, the electronic device can obtain a capacity detection result with an error within 0.1% through the target production line data, the capacity classification model and the capacity value prediction model. Therefore, compared with the traditional capacity detection scheme, the prediction accuracy can be significantly improved, and NG batteries (i.e. unqualified batteries) can be accurately detected, thereby achieving the purpose of preventing bad batteries from flowing out of the production line.

[0102] To more clearly illustrate the implementation of this application, please refer to Figure 3 , Figure 3 are schematic diagrams of application scenarios in certain embodiments of the present application. Figure 3 As shown, in the implementation mode of the present application, characteristic data related to battery capacity can be determined from all production line data of the battery through pre-performed theoretical analysis, correlation analysis (such as P value, correlation coefficient and main effect analysis, etc.), that is, target production line data related to at least one of the active lithium loss characteristics, internal resistance difference characteristics and capacity utilization characteristics.

[0103] In one example, an electronic device (or a control device within the electronic device) may perform a capacity detection method as shown in steps 01-03 above on each cell in the cell detection segmentation process of a lithium battery production line. In other words, in one example, the battery to be tested is a cell to be subjected to capacity detection in the cell detection segmentation process.

[0104] In one example, the target production line data includes a charging time parameter associated with a lithium loss characteristic.

[0105] In one example, the target production line data includes a charge relaxation voltage parameter associated with an internal resistance difference characteristic.

[0106] In one example, the target production line data includes temperature parameters related to capacity utilization characteristics.

[0107] In one example, the target production line data includes at least one of the above-mentioned charging time parameter, charging relaxation voltage parameter and temperature parameter.

[0108] In one example, the capacity classification model and the capacity value prediction model are both XGBoost models.

[0109] In one example, after receiving the target production line data of the battery to be tested, the capacity classification model can output one of two expected results based on the target production line data as the above-mentioned capacity category prediction result, such as a first result indicating that the capacity of the battery to be tested meets the production line process standards and thus the capacity of the battery to be tested is OK / qualified; and a second result indicating that the capacity of the battery to be tested exceeds the production line process standards and thus the capacity of the battery to be tested is NG / unqualified.

[0110] In one example, after receiving the target production line data of the battery to be tested, the capacity classification model may output a predicted capacity value of the battery to be tested according to the target production line data, that is, the above-mentioned capacity value prediction result.

[0111] See also Figure 4 In some implementations of the present application, the capacity category prediction results include qualified and unqualified, and the above step 02 includes:

[0112] 020: Determine the capacity category prediction result based on the target production line data and capacity classification model;

[0113] 021: When the capacity category prediction result is qualified, the capacity value prediction result is determined according to the target production line data of the battery to be tested and the capacity value prediction model.

[0114] The processing unit of the embodiment of the present application is configured to determine the capacity category prediction result based on the target production line data and the capacity classification model, and when the capacity category prediction result is qualified, determine the capacity value prediction result based on the target production line data and the capacity value prediction model of the battery to be tested.

[0115] Specifically, since the model calling operation involves corresponding resource consumption, in order to avoid ineffective resource consumption, in the implementation mode of the present application, the target production line data can be preferentially predicted through the capacity classification model to predict whether the capacity of the battery to be tested is qualified or unqualified.

[0116] Furthermore, if the capacity of the battery to be tested is predicted to be qualified, the capacity value prediction model is called to predict the target production line data to obtain the capacity value prediction result of the battery to be tested. On the contrary, if the capacity of the battery to be tested is predicted to be unqualified, the capacity detection process of the battery to be tested is directly terminated.

[0117] For illustration, please refer again Figure 3 Specifically, when the capacity classification prediction result is 1 (i.e. qualified), the target production line data will be input into the "output model 2" (i.e. capacity value prediction model) so that the capacity value prediction model predicts the capacity of the battery to be tested based on the received target production line data.

[0118] Thus, in the implementation mode of the present application, when the capacity classification model outputs a qualified capacity category prediction result based on the target production line data, the capacity value prediction model can be called to determine the capacity value prediction result of the battery to be tested, thereby reducing the resource consumption of the model call and avoiding invalid resource consumption to a certain extent.

[0119] In some implementations of the present application, the capacity category prediction result includes qualified and unqualified, and then, the above step 03 includes:

[0120] When the capacity category prediction result is qualified and the capacity value prediction result is greater than or equal to a preset capacity threshold, determining that the capacity of the battery to be tested is qualified; and / or

[0121] When the capacity category prediction result is unqualified, or the capacity value prediction result is less than a preset capacity threshold, it is determined that the capacity of the battery to be tested is unqualified.

[0122] The processing unit of the embodiment of the present application is configured to determine that the capacity of the battery to be tested is qualified when the capacity category prediction result is qualified and the capacity value prediction result is greater than or equal to a preset capacity threshold; and / or determine that the capacity of the battery to be tested is unqualified when the capacity category prediction result is unqualified or the capacity value prediction result is less than the preset capacity threshold.

[0123] Specifically, to clearly illustrate the implementation methods of this application, please refer to Figure 3 That is, in Figure 2In an example shown, when the capacity classification model predicts the target production line data and predicts that the capacity of the battery to be tested is qualified (i.e., the capacity classification prediction result is 1), and the capacity value prediction model predicts the target production line data and predicts that the capacity value of the battery to be tested is greater than or equal to the preset capacity threshold (i.e., the capacity value prediction result is 1), it can be determined that the capacity of the battery to be tested meets the production line process standards and the capacity of the battery to be tested is OK / qualified.

[0124] And as Figure 3 In another example shown, when the capacity classification model predicts the target production line data and predicts that the capacity of the battery to be tested is unqualified (i.e., the capacity classification prediction result is 0), or the capacity value prediction model predicts the target production line data and predicts that the capacity value of the battery to be tested is less than a preset capacity threshold (i.e., the capacity value prediction result is 0), it can be determined that the capacity of the battery to be tested does not meet the production line process standards, and the capacity of the battery to be tested is NG / unqualified.

[0125] Thus, in the embodiments of the present application, the capacity of the battery to be tested can be determined to be qualified when the capacity category prediction result is qualified and the capacity value prediction result is greater than or equal to the preset capacity threshold, and / or the capacity of the battery to be tested can be determined to be unqualified when the capacity category prediction result is unqualified or the capacity value prediction result is less than the preset capacity threshold, thereby ensuring the reliability and accuracy of the battery capacity detection results to a certain extent.

[0126] See also Figure 5 In certain embodiments of the present application, the method further comprises:

[0127] 04: Obtain the capacity label of the first battery sample and various first production line data;

[0128] 05: Process the capacity label and the first production line data to determine the correlation analysis result between the capacity label and the first production line data;

[0129] 06: Determine target production line data among the multiple first production line data according to the correlation analysis processing results of each type of first production line data.

[0130] The transceiver unit of the embodiment of the present application is configured to obtain the capacity label of the first battery sample and multiple first production line data. The processing unit of the embodiment of the present application is configured to process the capacity label and the first production line data, determine the correlation analysis result between the capacity label and the first production line data, and determine the target production line data among the multiple first production line data according to the correlation analysis processing result of each first production line data.

[0131] It is understandable that lithium-ion power batteries are a complex system, and the factors affecting capacity are nonlinear. At the same time, the capacity detection process has a significant impact on the capacity of the capacity distribution. Therefore, in order to ensure the accuracy of the above capacity classification model and capacity value detection model, the factors affecting capacity can be fully analyzed.

[0132] It is also understandable that, since the production line process data and laboratory data dimensions are different, the factors affecting capacity can be characterized and quantified to extract capacity classification models and capacity value detection model data that can be used in actual production lines. Therefore, the factors that characterize active lithium loss, internal resistance differences, and factors that affect capacity differences can be converted into production line data features, such as charging time, charging relaxation voltage, and temperature data.

[0133] Based on this, in the implementation mode of the present application, the electronic device can obtain the capacity label of the first battery sample (i.e., the actual capacity of the first battery sample), and obtain multiple first production line data of the first battery sample. Then, the capacity label and the first production line data are processed such as P value, correlation coefficient and main effect analysis to determine the correlation analysis results between the capacity label and the first production line data. Finally, according to the correlation analysis processing results of each first production line data, the data with a higher correlation with the capacity label among the multiple first production line data is determined, and these data are used as target production line data for use in actual capacity detection.

[0134] For example, when it is determined through P value, correlation coefficient and main effect analysis that the capacity label has a strong direct correlation with the charge relaxation voltage, the electronic device can obtain the charge relaxation voltage of the battery to be tested in an actual capacity detection environment as a basis for capacity detection of the battery to be tested.

[0135] Thus, in the embodiment of the present application, the capacity label and multiple first production line data of the first battery sample can be obtained. Then, the capacity label and the first production line data are processed to determine the correlation analysis result between the capacity label and the first production line data, and according to the correlation analysis processing result of each first production line data, the target production line data among the multiple first production line data is determined, so that the correlation between the target production line data and the battery capacity is guaranteed.

[0136] See also Figure 6 In certain embodiments of the present application, step 04 includes:

[0137] 040: Obtain a capacity label of a first battery sample and a plurality of second production line data;

[0138] 041: Perform data cleaning processing on the second production line data to determine the first production line data, wherein the data cleaning processing includes data length alignment processing and / or outlier elimination processing.

[0139] The processing unit of the embodiment of the present application is configured to obtain the capacity label of the first battery sample and multiple second production line data, and perform data cleaning processing on the second production line data to determine the first production line data, and the data cleaning processing includes data length alignment processing and / or outlier removal processing.

[0140] Specifically, in order to avoid the interference of data noise in the process of determining the target production line data, in the implementation mode of the present application, when the original production line data is obtained, or when a plurality of second production line data are obtained, data cleaning processing can be performed on each type of second production line data, and then the cleaned second production line data can be determined as the first production line data.

[0141] In one example, the electronic device may align the collected raw data (i.e., the second production line data) set through barcodes, and then remove abnormal values ​​such as missing values ​​and values ​​exceeding process standards in the raw data, thereby completing the data cleaning operation.

[0142] Thus, in the embodiment of the present application, a data cleaning operation may be performed on the acquired second production line data to obtain the first production line data, thereby ensuring the validity and reliability of the first production line data and the target production line data.

[0143] In certain embodiments of the present application, the correlation analysis result includes a first sub-result, and further, the above step 05 includes:

[0144] The capacity label and the first production line data are analyzed and processed using a scatter matrix diagram to determine a first sub-result.

[0145] The processing unit of the implementation mode of the present application is configured to perform scatter matrix analysis on the capacity label and the first production line data to determine the first sub-result.

[0146] Specifically, in the implementation manner of the present application, the electronic device can analyze the correlation between each capacity label and each first production line data through a matrix scatter plot (Matrix ScatterPlot), or in other words, analyze whether there is a corresponding functional relationship between the first production line data and the capacity.

[0147] It can be understood that the matrix scatter plot can show the pairwise correlation of multidimensional data, which can be used to compare aggregated data across categories and summarize and analyze existing data for predictive analysis. Using the scatter plot matrix to draw scatter plots between each variable at the same time, the main correlation between multiple variables can be quickly discovered.

[0148] Thus, in the embodiment of the present application, the correlation between the capacity label and each first production line data can be analyzed through a matrix scatter plot.

[0149] In certain embodiments of the present application, the correlation analysis result includes a second sub-result, and further, the above step 05 includes:

[0150] The assumption that there is a correlation between the capacity label and the first production line data is verified to obtain a second sub-result.

[0151] The processing unit of the embodiment of the present application is configured to verify the assumption that there is a correlation between the capacity tag and the first production line data to obtain a second sub-result.

[0152] Specifically, in the embodiment of the present application, the electronic device may calculate a P value to analyze the correlation between the capacity label and the first production line data.

[0153] It can be understood that the P value can be understood as the probability of a result more extreme than the sample observation result when the null hypothesis is true.

[0154] It can also be understood that if the P value is small, it means that the probability of the occurrence of the original hypothesis is very small. It can be considered that the higher the probability that the hypothesis "there is a correlation between the capacity label and the first production line data" is not true, the more sufficient the reason for rejecting the hypothesis.

[0155] For example, if the data of a first production line is x and the capacity label is y, then for the assumption that there is a correlation between x and y, if the p-value is ≤ 0.05, it indicates that x has a significant impact on y.

[0156] Thus, in the embodiment of the present application, the assumption that there is a correlation between the capacity tag and the first production line data can be verified, thereby analyzing the correlation between each capacity tag and each first production line data.

[0157] In certain embodiments of the present application, the correlation analysis result includes a third sub-result, and further, the above step 05 includes:

[0158] The correlation coefficient between the capacity label and the first production line data is calculated to obtain the third sub-result.

[0159] The processing unit of the implementation mode of the present application is configured to calculate the correlation coefficient between the capacity label and the first production line data to obtain a third sub-result.

[0160] Specifically, in the implementation manner of the present application, the correlation between the capacity label and the first production line data can be determined by calculating the correlation coefficient.

[0161] It is understandable that if the correlation coefficient r between a first production line data x and a capacity label y is yxThe closer it is to 1, the greater the impact of the first production line data x on the capacity label y, and the more obvious the correlation between the first production line data x and the capacity label y. It can also be understood that |r yx |≤1.

[0162] Thus, in the embodiment of the present application, the correlation coefficient between the capacity tag and the first production line data may be calculated to analyze the correlation between the capacity tag and the first production line data.

[0163] In certain embodiments of the present application, the correlation analysis result includes a fourth sub-result, and further, the above step 05 includes:

[0164] The capacity label and the first production line data were subjected to main effect analysis to obtain the fourth sub-result.

[0165] The processing unit of the implementation mode of the present application is configured to perform main effect analysis processing on the capacity label and the first production line data to obtain a fourth sub-result.

[0166] Specifically, in an embodiment of the present application, the electronic device may perform main effect analysis on the capacity label and each first production line data to determine the significance of the impact of each first production line data on the capacity, thereby determining the correlation between the capacity label and each first production line data.

[0167] It can be understood that the main effect refers to the variation caused by different levels of a factor in the experiment. The main effect analysis can evaluate the impact of a single factor on the dependent variable at different levels. For example, when there are two factors A and B, factor A has three levels (A1, A2, A3), and factor B has two levels (B1, B2), the main effect analysis can be used to evaluate the impact of factor A on the dependent variable at different levels. Therefore, in the embodiment of the present application, the impact of the first production line data on the capacity label can be evaluated by taking different values ​​of the first production line data, thereby determining the correlation analysis result between the first production line data and the capacity label, that is, the fourth sub-result mentioned above.

[0168] In one example, a main effect analysis can be performed to determine the difference (y) between the capacity mean values ​​y corresponding to the high level and the low level of the first production line data. high -y low ) max . Further, (y high -y low ) max The larger the value is, the greater the impact of the first production line data on the capacity is. high -y low ) max is the main effect of capacity.

[0169] Thus, in the implementation mode of the present application, the significance of the impact of each first production line data on the capacity can be determined by the main effect analysis method, thereby determining the correlation between the capacity label and each first production line data.

[0170] In certain embodiments of the present application, the correlation analysis result includes multiple sub-results, and the above step 06 includes:

[0171] Among the multiple types of first production line data, first production line data in which each sub-result satisfies corresponding preset conditions is determined as target production line data.

[0172] The processing unit in the embodiment of the present application is configured to determine, among the multiple types of first production line data, first production line data in which each sub-result satisfies the corresponding preset condition as target production line data.

[0173] Specifically, in the implementation manner of the present application, for each type of first production line data, the electronic device can analyze the correlation between the first production line data and the capacity label through a plurality of different correlation analysis methods. Therefore, for each type of first production line data, each type of first production line data corresponds to a plurality of correlation analysis results, or in other words, each type of first production line data corresponds to a correlation analysis result composed of a plurality of sub-results.

[0174] In one example, the electronic device can analyze the correlation between each first production line data and the capacity label through four methods: P value verification, main effect analysis, matrix scatter plot analysis, and correlation coefficient calculation. In other words, the correlation analysis result of any first production line data includes the above-mentioned first sub-result, second sub-result, third sub-result, and fourth sub-result.

[0175] Furthermore, the electronic device may construct a correlation analysis quadrant diagram according to the first sub-result, the second sub-result, the third sub-result and the fourth sub-result of each type of first production line data to determine the target production line data among the multiple types of first production line data.

[0176] In one example, let the second sub-result be the above P value, the second sub-result be the above r, and the third sub-result be the above r. yx The fourth result is the above (y high -y low ) max , then: there is a functional relationship between the first production line data and the capacity label, P value ≤ 0.05, 0.4 ≤ | r yx |≤1, and (y high -y low ) max If it is greater than or equal to a preset value, the first production line data is determined as the target production line data.

[0177] For example, the first production line data acquired by the electronic device includes the charging time T of a plurality of first battery samples. C and the discharge duration T of the plurality of first battery samples D In the case of charging time T C There is a functional relationship with the capacity label Y, and the charging time T C The corresponding P value is ≤0.05, and the charging time is T C The correlation coefficient r between the capacity label Y yx The absolute value of is greater than or equal to 0.4 and less than or equal to 1, and the charging time is T C The main effect analysis results (y high -y low ) max is greater than or equal to the preset value, and the discharge time T D There is no functional relationship with the capacity label Y, or the discharge time T D The corresponding P value is greater than 0.05, and the discharge time is T D The correlation coefficient r between the capacity label Y yx The absolute value is less than 0.4, and the discharge time is T D The main effect analysis results (y high -y low ) max If the charging time is less than the preset value, C and discharge time T D Medium, charging time T C Can be used as target production line data.

[0178] Thus, in the embodiment of the present application, among a plurality of first production line data, the first production line data in which each sub-result satisfies the corresponding preset conditions can be determined as the target production line data, thereby ensuring the reliability and validity of the target production line data.

[0179] See also Figure 7 In certain embodiments of the present application, the training step of the capacity classification model includes:

[0180] 21: Obtain a first data set, wherein the first data set includes target production line data and capacity category labels of the second battery sample;

[0181] 22: Perform model training based on the first data set to determine the capacity classification model.

[0182] The transceiver unit of the embodiment of the present application is configured to obtain a first data set, the first data set including target production line data and capacity category labels of the second battery sample. The processing unit is configured to perform model training according to the first data set to determine a capacity classification model.

[0183] Specifically, in an embodiment of the present application, the electronic device can obtain the target production line data of the second battery sample and the capacity category label of the second battery sample to construct a data set for training the capacity classification model, and perform supervised model training through the first data set to obtain a capacity classification model that can output capacity category prediction results based on the input of the target production line data.

[0184] In one example, when the actual capacity of the second battery sample meets the expected production line process standard, the capacity category label of the second battery sample may be set to 1, and when the actual capacity of the second battery sample exceeds the expected production line process standard, the capacity category label of the second battery sample may be set to 0.

[0185] Thus, in an embodiment of the present application, a first data set may be obtained and model training may be performed based on the first data set to determine a capacity classification model, so as to obtain a capacity classification model that can output capacity category prediction results based on the input of target production line data through supervised training.

[0186] See also Figure 8 In certain embodiments of the present application, step 22 includes:

[0187] 220: Determine a first training set and a first test set according to the first data set;

[0188] 221: training a first model according to a first training set;

[0189] 222: Determine the classification accuracy of the trained first model according to the first test set;

[0190] 223: When the classification accuracy is greater than or equal to the first threshold, determine the trained first model as a capacity classification model.

[0191] The processing unit of the embodiment of the present application is configured to determine a first training set and a first test set based on a first data set, train a first model based on the first training set, determine the classification accuracy of the trained first model based on the first test set, and determine the trained first model as a capacity classification model when the classification accuracy is greater than or equal to a first threshold.

[0192] Specifically, in an embodiment of the present application, after obtaining the first data set, the electronic device may divide the first data set to divide a part of the first data set into a training set and the remaining part into a test set, thereby obtaining a first training set and a first test set. Then, the first model (i.e., the capacity classification model that has not yet been trained) is trained through the first training set to update parameters such as bias and weight in the first model. Then, the first model after completing one training is performance tested through the first test set to test the probability that the capacity prediction result predicted by the first model based on the target production line data of the second battery sample is the capacity category label of the second battery sample, that is, the classification accuracy.

[0193] Finally, if the classification accuracy is higher than or equal to the preset first threshold, it can be confirmed that the trained first model has met expectations, and the trained first model is used as the capacity classification model. On the contrary, if the classification accuracy is less than the first threshold, it can be confirmed that the trained first model has not met expectations, and the first training set and the first test set can be re-divided according to the first data set to start a new round of training, thereby realizing the iterative update of the first model.

[0194] In one example, the electronic device sets the first training set and the first test set based on a cross-validation approach.

[0195] Thus, in an embodiment of the present application, a first training set and a first test set can be determined based on a first data set, and a first model can be trained based on the first training set, and a classification accuracy of the trained first model can be determined based on the first test set, and when the classification accuracy is greater than or equal to a first threshold, the trained first model is determined as a capacity classification model, thereby achieving training of the first model.

[0196] In certain embodiments of the present application, the method further comprises:

[0197] Adjust the hyperparameters of the first model, wherein the hyperparameters include at least one of the number of tree models, the learning rate, the maximum depth of the decision tree, the regularization term weight, the random seed, the random sampling ratio of samples, and the random sampling ratio of features.

[0198] Specifically, in the embodiment of the present application, the first model is an XGBoost model. Therefore, before training the first model, one or more of the seven hyperparameters of the first model, including the number of tree models (n_es-timators), learning rate (learning_rate), maximum depth of decision tree (max_depth), weight of regularization term (alpha), random seed (seed), random sampling ratio of samples (subsample) and random sampling ratio of features (colsam-ple_bytree), can be adjusted to improve the training effect of the first model.

[0199] Thus, in the implementation manner of the present application, the hyperparameters of the first model may be adjusted before the training of the first model begins, so as to improve the training effect of the first model to a certain extent.

[0200] See also Fig. 9 In certain embodiments of the present application, the training step of the capacity value prediction model includes:

[0201] 31: Acquire a second data set, wherein the second data set includes target production line data and a capacity value label of a third battery sample;

[0202] 32: Perform model training based on the second data set to determine a capacity value prediction model.

[0203] The transceiver unit of the embodiment of the present application is configured to obtain a second data set, the second data set including target production line data and capacity value labels of the third battery sample. The processing unit is configured to perform model training according to the second data set to determine a capacity value prediction model.

[0204] Specifically, in an embodiment of the present application, the electronic device can obtain the target production line data of the third battery sample and the capacity value label of the third battery sample to construct a data set for training the capacity classification model, and perform supervised model training through the first data set to obtain a capacity value prediction model that can output capacity value prediction results based on the input of the target production line data.

[0205] Thus, in an embodiment of the present application, a first data set may be obtained and model training may be performed based on the first data set to determine a capacity classification model, so as to obtain a capacity value prediction model that can output a capacity value prediction result based on the input of target production line data through supervised training.

[0206] See also Fig.10 In some embodiments of the present application, step 32 includes:

[0207] 320: Determine a second training set and a second test set according to the second data set;

[0208] 321: training a second model according to a second training set;

[0209] 322: Determine a prediction error parameter of the trained second model according to the second test set;

[0210] 323: When the prediction error parameter is less than or equal to the second threshold, determine the trained second model as the capacity value prediction model.

[0211] The processing unit of the embodiment of the present application is configured to determine a second training set and a second test set based on a second data set, and to train a second model based on the second training set, and to determine a prediction error parameter of the trained second model based on the second test set, and to determine the trained second model as a capacity value prediction model when the prediction error parameter is less than or equal to a second threshold.

[0212] Specifically, in an embodiment of the present application, the electronic device may divide the second data set after obtaining the second data set, so as to divide a part of the second data set into a training set and the remaining part into a test set, thereby obtaining a second training set and a second test set. Then, the second model (i.e., the capacity value prediction model that has not yet been trained) is trained through the second training set to update the parameters such as bias and weight in the second model. Then, the second model after completing one training is performance tested through the second test set to test the error between "the capacity value prediction result predicted by the second model based on the target production line data of the third battery sample" and "the capacity value label of the third battery sample", that is, the prediction error parameter.

[0213] Finally, if the prediction error parameter is less than or equal to the preset second threshold, it can be confirmed that the trained second model has met expectations, and the trained second model is used as the capacity value prediction model. On the contrary, if the prediction error parameter is greater than the second threshold, it can be confirmed that the trained second model has not met expectations, and the second training set and the second test set can be re-divided according to the second data set to start a new round of training, thereby realizing the iterative update of the second model.

[0214] In one example, the electronic device sets the second training set and the second test set based on a cross-validation approach.

[0215] In one example, the prediction error parameter is calculated using a mean squared error (MSE) formula.

[0216] In one example, the trained second model (i.e., the capacity value prediction model) predicts the target production line data of a battery sample, and the prediction error between the obtained capacity value prediction result and the capacity value label of the battery sample is within 0.1%. It can be understood that compared with the traditional capacity detection solution, the capacity prediction accuracy based on the capacity value prediction model has been significantly improved, and it can reliably guarantee the detection of NG batteries and achieve the purpose of preventing the outflow of defective batteries from the production line.

[0217] Thus, in an embodiment of the present application, a second training set and a second test set can be determined based on the second data set, and a second model can be trained based on the second training set, and a prediction error parameter of the trained second model can be determined based on the second test set, and when the prediction error parameter is less than or equal to a second threshold, the trained second model is determined as a capacity value prediction model, thereby realizing the training of the second model.

[0218] In certain embodiments of the present application, the method further comprises:

[0219] Adjust the hyperparameters of the second model, wherein the hyperparameters include at least one of the number of tree models, the learning rate, the maximum depth of the decision tree, the regularization term weight, the random seed, the random sampling ratio of samples, and the random sampling ratio of features.

[0220] The processing unit of the embodiment of the present application is configured to adjust the hyperparameters of the second model, and the hyperparameters include at least one of the number of tree models, learning rate, maximum depth of decision tree, regularization term weight, random seed, random sampling ratio of samples and random sampling ratio of features.

[0221] Specifically, in the embodiment of the present application, the second model is an XGBoost model. Therefore, before training the second model, one or more of the seven hyperparameters of the second model, including the number of tree models (n_es-timators), learning rate (learning_rate), maximum depth of decision tree (max_depth), weight of regularization term (alpha), random seed (seed), random sampling ratio of samples (subsample) and random sampling ratio of features (colsam-ple_bytree), can be adjusted to improve the training effect of the second model.

[0222] Thus, in the implementation manner of the present application, the hyperparameters of the second model may be adjusted before the training of the second model begins, so as to improve the training effect of the second model to a certain extent.

[0223] Please refer again Figure 3In one example, the electronic device may determine the target production line data from the multiple first production line data based on the multiple first production line data of the first battery sample, collect the target production line data of the battery to be tested, perform data cleaning on the received target production line data, and then input the cleaned target production line data into the trained model 1 (i.e., the capacity classification model) and model 2 (i.e., the capacity value prediction model) in sequence.

[0224] Next, model 1 outputs the predicted label, and the battery data with the predicted label 1 continues to be input into model 2.

[0225] Next, model 2 outputs the capacity prediction result. If the predicted capacity meets the production line process standard, the prediction result label output is 1; if the predicted capacity exceeds the production line process standard, the predicted capacity label output is 0.

[0226] Finally, the battery whose overall prediction result prediction label is 1 (that is, the output label of the capacity classification model and the output label of the capacity value prediction model are both 1) is regarded as an OK battery / qualified battery, and the battery whose overall prediction result prediction label is 0 (that is, the output label of the capacity classification model or the output label of the capacity value prediction model is 0) is regarded as an NG battery / unqualified battery.

[0227] In certain embodiments of the present application, the target production line data includes a plurality of production line parameters, and the method further includes:

[0228] According to the battery capacity prediction model, the contribution of each production line parameter to the battery capacity prediction model is determined, wherein the battery capacity prediction model is pre-trained and can predict the capacity value of the battery to be tested based on the target production line data of the battery to be tested.

[0229] The processing unit of the embodiment of the present application is configured to determine the contribution of each production line parameter to the battery capacity prediction model based on the battery capacity prediction model. The battery capacity prediction model is pre-trained and can predict the capacity value of the battery to be tested based on the target production line data of the battery to be tested.

[0230] Specifically, in one example, the capacity classification model and the capacity value prediction model are both models built based on XGBoost. Accordingly, in order to explain the role of each parameter in the target production line data for model prediction, in the embodiment of the present application, a battery capacity prediction model based on XGBoost can be constructed, and through the battery capacity prediction model, the contribution of each production line parameter in the target production line data to the battery capacity prediction model can be evaluated, thereby explaining the impact of each parameter on capacity prediction.

[0231] In one example, after the contribution of each production line parameter to the battery capacity prediction model is obtained, the contribution of each production line parameter may be normalized to eliminate dimensional differences.

[0232] In one example, the higher the contribution of the production line parameters, the greater the effect of the production line parameters on the capacity prediction of the battery capacity prediction model, thereby achieving the interpretability of the model built based on XGBoost (such as the above-mentioned capacity classification model and the above-mentioned capacity value prediction model).

[0233] In one example, the production line parameters with the highest contribution may be analyzed, such as correlation coefficient calculation and main effect analysis, to determine the influencing factors of the production line parameters so as to identify the iteration direction of model optimization.

[0234] Thus, in the embodiment of the present application, the contribution of each production line parameter to the battery capacity prediction model can be determined based on the battery capacity prediction model to explain the contribution of each production line parameter to the capacity classification model and the capacity value prediction model.

[0235] The embodiment of the present application further provides an electronic device, which includes the above-mentioned control device.

[0236] The embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by one or more processors, the above-mentioned battery capacity detection method is implemented.

[0237] In the description of this specification, the descriptions with reference to the terms "specifically", "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 representations 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.

[0238] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes 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 not be performed in 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 technicians in the technical field to which the embodiments of the present application belong.

[0239] 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 method for detecting battery capacity, characterized in that: include: Acquire target production line data of the battery to be tested, wherein the target production line data is related to at least one of an active lithium loss characteristic, an internal resistance difference characteristic, and a capacity utilization characteristic of the battery to be tested; Determine a capacity category prediction result and a capacity value prediction result of the battery to be tested according to target production line data of the battery to be tested, and a pre-built capacity classification model and a capacity value prediction model; The capacity of the battery to be tested is determined according to the capacity category prediction result and the capacity value prediction result, and the capacity detection is completed.

2. The method according to claim 1, characterized in that The capacity category prediction result includes qualified and unqualified, and the capacity category prediction result and the capacity value prediction result of the battery to be tested are determined according to the target production line data of the battery to be tested, and the pre-built capacity classification model and capacity value prediction model, including: Determining the capacity category prediction result according to the target production line data and the capacity classification model; When the capacity category prediction result is qualified, the capacity value prediction result is determined according to the target production line data of the battery to be tested and the capacity value prediction model.

3. The method according to claim 1, characterized in that The capacity category prediction result includes qualified and unqualified, and the capacity of the battery to be tested is determined according to the capacity category prediction result and the capacity value prediction result to complete the capacity detection, including: When the capacity category prediction result is qualified and the capacity value prediction result is greater than or equal to a preset capacity threshold, determining that the capacity of the battery to be tested is qualified; and / or When the capacity category prediction result is unqualified, or the capacity value prediction result is less than the preset capacity threshold, it is determined that the capacity of the battery to be tested is unqualified.

4. The method according to claim 1, characterized in that The method further comprises: Obtaining a capacity label and a plurality of first production line data of a first battery sample; Processing the capacity tag and the first production line data to determine a correlation analysis result between the capacity tag and the first production line data; The target production line data among the plurality of first production line data are determined according to the correlation analysis processing result of each type of the first production line data.

5. The method according to claim 4, characterized in that The step of obtaining the capacity label of the first battery sample and a plurality of first production line data includes: Acquire the capacity label and a plurality of second production line data of the first battery sample; The second production line data is cleaned to determine the first production line data, wherein the data cleaning includes data length alignment and / or outlier elimination.

6. The method according to claim 4, characterized in that The correlation analysis result includes a first sub-result, and the processing of the capacity tag and the first production line data to determine the correlation analysis result of the capacity tag and the first production line data includes: The capacity label and the first production line data are subjected to scatter matrix analysis to determine the first sub-result.

7. The method according to claim 4, characterized in that The correlation analysis result includes a second sub-result, and the processing of the capacity tag and the first production line data to determine the correlation analysis result of the capacity tag and the first production line data includes: The assumption that there is a correlation between the capacity label and the first production line data is verified to obtain the second sub-result.

8. The method according to claim 4, characterized in that The correlation analysis result includes a third sub-result, and the processing of the capacity tag and the first production line data to determine the correlation analysis result of the capacity tag and the first production line data includes: The correlation coefficient between the capacity label and the first production line data is calculated to obtain the third sub-result.

9. The method according to claim 4, characterized in that The correlation analysis result includes a fourth sub-result, wherein the processing of the capacity tag and the first production line data to determine the correlation analysis result of the capacity tag and the first production line data includes: Main effect analysis is performed on the capacity label and the first production line data to obtain the fourth sub-result.

10. The method according to claim 4, characterized in that The correlation analysis result includes a plurality of sub-results, and the determining the target production line data among the plurality of first production line data according to the correlation analysis processing result of each type of the first production line data includes: Among the plurality of first production line data, first production line data in which each of the sub-results satisfies a corresponding preset condition is determined as the target production line data.

11. The method according to claim 1, characterized in that: The training steps of the capacity classification model include: Acquire a first data set, wherein the first data set includes target production line data and capacity category labels of a second battery sample; Model training is performed according to the first data set to determine the capacity classification model.

12. The method according to claim 11, characterized in that The performing model training according to the first data set to determine the capacity classification model includes: Determine a first training set and a first test set according to the first data set; Training a first model according to the first training set; Determining the classification accuracy of the first model after training according to the first test set; When the classification accuracy is greater than or equal to a first threshold, the trained first model is determined as the capacity classification model.

13. The method according to claim 12, characterized in that The method further comprises: Adjust the hyperparameters of the first model, wherein the hyperparameters include at least one of the number of tree models, learning rate, maximum depth of decision tree, regularization term weight, random seed, random sampling ratio of samples, and random sampling ratio of features.

14. The method according to claim 1, characterized in that The training steps of the capacity value prediction model include: Acquire a second data set, wherein the second data set includes target production line data and capacity value labels of a third battery sample; Model training is performed according to the second data set to determine the capacity value prediction model.

15. The method according to claim 14, characterized in that The performing model training according to the second data set to determine the capacity value prediction model includes: Determine a second training set and a second test set according to the second data set; Training a second model based on the second training set; Determining a prediction error parameter of the trained second model according to the second test set; When the prediction error parameter is less than or equal to a second threshold, the trained second model is determined as the capacity value prediction model.

16. The method according to claim 15, characterized in that The method further comprises: Adjust the hyperparameters of the second model, wherein the hyperparameters include at least one of the number of tree models, learning rate, maximum depth of decision tree, regularization term weight, random seed, random sampling ratio of samples, and random sampling ratio of features.

17. The method according to claim 1, characterized in that The target production line data includes a plurality of production line parameters, and the method further includes: According to the battery capacity prediction model, the contribution of each of the production line parameters to the battery capacity prediction model is determined, wherein the battery capacity prediction model is pre-trained and can predict the capacity value of the battery to be tested according to the target production line data of the battery to be tested.

18. A control device, characterized in that: including a transceiver unit and a processing unit; The transceiver unit is configured to obtain target production line data of the battery to be tested, wherein the target production line data is related to the capacity performance characteristics of the battery to be tested; The processing unit is configured to determine the capacity category prediction result and the capacity value prediction result of the battery to be tested according to the target production line data of the battery to be tested, and a pre-built capacity classification model and a capacity value prediction model, and determine the capacity of the battery to be tested according to the capacity category prediction result and the capacity value prediction result to complete the capacity detection.

19. An electronic device, characterized in that: Comprising the control device as claimed in claim 18.

20. 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, the method according to any one of claims 1 to 17 is implemented.

Images