Model training method, battery testing method, device, equipment, medium and product

Abstract

The embodiment of the invention provides a model training method, a battery testing method, a device, equipment, a medium and a product. The model training method comprises the following steps: acquiring electrical performance test data for a plurality of batteries; obtaining a plurality of training samples based on the electrical performance test data of each battery and a preset test item set; on the basis of the plurality of training samples, continuously executing training operation on the battery performance test model until a preset training ending condition is met, and obtaining a trained battery performance test model; wherein the training operation comprises the following steps: for each training sample, based on each first test value in the training sample, predicting the full charge and discharge performance of the battery corresponding to the training sample through a battery performance test model, and obtaining prediction data of a full charge and discharge performance test item; and determining training loss based on the difference between the second test value corresponding to each training sample and the prediction data, and adjusting model parameters of the battery performance test model based on the training loss.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and more specifically, to a model training method, a battery testing method, an apparatus, a device, a medium, and a product. Background Art

[0002] In existing technologies, with the rapid development of electric vehicles and energy storage technologies, the demand for lithium batteries continues to grow. To ensure the stability and reliability of lithium battery performance in actual use, battery electrical performance testing has become a crucial step in the production process. Battery electrical performance testing includes shallow charge and discharge performance testing and full charge and discharge performance testing. The full charge and discharge performance testing of the battery takes longer than the shallow charge and discharge performance testing. Due to the longer test time of the full charge and discharge performance testing of the battery, the battery testing efficiency is lower. Summary of the Invention

[0003] In response to the shortcomings of existing methods, the present disclosure proposes a model training method, a battery testing method, an apparatus, a device, a computer-readable storage medium, and a computer program product to solve the problem of how to improve battery testing efficiency.

[0004] In a first aspect, the present disclosure provides a model training method, comprising: Acquire electrical performance test data for a plurality of batteries, wherein the electrical performance test data for each battery includes shallow charge and discharge performance test data and full charge and discharge performance test data of the battery; Based on the electrical performance test data of each battery and a preset test item set, a plurality of training samples are obtained, wherein the test item set includes a plurality of shallow charge and discharge performance test items and a full charge and discharge performance test item, and each training sample includes a first test value of a battery corresponding to at least one shallow charge and discharge performance test item and a second test value of the battery corresponding to the full charge and discharge performance test item; Based on multiple training samples, the battery performance test model is continuously trained until a preset training end condition is met, thereby obtaining a trained battery performance test model. The training operation includes: For each training sample, based on the first test values ​​in the training sample, the full charge and discharge performance of the battery corresponding to the training sample is predicted by the battery performance test model to obtain the predicted data of the full charge and discharge performance test item; based on the difference between the second test value corresponding to each training sample and the predicted data, the training loss is determined, and the model parameters of the battery performance test model are adjusted based on the training loss.

[0005] In one embodiment, a preset model set is obtained, where the model set includes a plurality of battery performance test models; Based on multiple training samples, the battery performance test model is continuously trained until a preset training end condition is met, thereby obtaining a trained battery performance test model, including: Based on the multiple training samples, the multiple battery performance test models in the model set are continuously trained until a preset training end condition is met, thereby obtaining the multiple trained battery performance test models; The model performance of the multiple battery performance test models after training is tested respectively, and the model with the best performance among the multiple battery performance test models after training is determined as the trained battery performance test model.

[0006] In one embodiment, the plurality of training samples are divided into a training set and a test set; Based on multiple training samples, training operations are continuously performed on multiple battery performance test models in the model set, including: Based on multiple training samples in the training set, training operations are continuously performed on multiple battery performance test models respectively; Among them, the trained battery performance test model is determined by the following method: Based on multiple training samples in the test set and preset model performance test indicators, the model performance of multiple trained battery performance test models is tested respectively to obtain the index value of the model performance test indicator corresponding to each trained battery performance test model; Based on the indicator values ​​corresponding to the trained battery performance test models, the model with the best model performance among the trained battery performance test models is determined as the trained battery performance test model.

[0007] In one embodiment, for each battery, the electrical performance test data of the battery is obtained by: Acquiring multiple original electrical performance test data obtained by performing electrical performance tests on the battery within a preset time period; For each test item in the test item set, extracting a test value of the test item from multiple original electrical performance test data of the battery based on a preset test value extraction method for the test item; Based on the test value of each test item in the extracted test item set, the electrical performance test data of the battery is determined.

[0008] In one embodiment, the multiple shallow charge and discharge performance test items include at least two of temperature, voltage, stage discharge capacity, stage discharge energy, stage discharge pressure difference, charging energy, and charging capacity; the full charge and discharge performance test item includes the dynamic pressure difference at the end of discharge; For each test item of temperature and voltage, the test value extraction method corresponding to the test item includes extracting the maximum value or minimum value of the test value of the test item in the original electrical performance test data for multiple times, or extracting the difference between the maximum value and the minimum value of the test value of the test item in the original electrical performance test data for multiple times; For each test item of stage discharge capacity, stage discharge energy, charging energy, and charging capacity, a method for extracting a test value corresponding to the test item includes extracting a cumulative value of the test value of the test item from multiple original electrical performance test data; For each test item in the stage discharge pressure difference and the discharge end dynamic pressure difference, the test value extraction method corresponding to the test item includes extracting the difference between the maximum and minimum test values ​​of the test item in multiple original electrical performance test data.

[0009] In one embodiment, for each battery, the electrical performance test data of the battery is obtained by: Acquiring electrical performance test data for the battery in each of a plurality of preset time periods; The electrical performance test data of the battery in the first time period among the multiple preset sections is used as the electrical performance test data of the battery.

[0010] In one embodiment, based on the electrical performance test data of each battery and a preset test item set, a plurality of training samples are obtained, including: For each battery, if it is determined based on the electrical performance test data and the test item set of the battery that the test value corresponding to any test item in the test item set of the battery is missing or the test value is abnormal, the electrical performance test data of the battery is deleted; A plurality of training samples are obtained based on the first test value of each retained battery corresponding to each shallow charge and discharge performance test item in the test item set and the second test value corresponding to the full charge and discharge performance test item.

[0011] In one embodiment, a plurality of training samples are obtained based on the first test value of each battery corresponding to each shallow charge and discharge performance test item in the test item set and the second test value corresponding to the full charge and discharge performance test item, including: For each shallow charge and discharge performance test item among a plurality of shallow charge and discharge performance test items, determining a correlation between the shallow charge and discharge performance test item and a full charge and discharge performance test item based on first test data and second test data corresponding to the test item; wherein the first test data includes a first test value of each retained battery corresponding to the shallow charge and discharge performance test item, and the second test data includes a second test value of the full charge and discharge performance test item; For each shallow charge and discharge performance test item, if the correlation corresponding to the shallow charge and discharge performance test item is less than a preset correlation threshold, the shallow charge and discharge performance test item is deleted from the test item set; For each retained battery, a training sample is obtained based on the battery corresponding to the first test data and the battery corresponding to the second test data.

[0012] In a second aspect, the present disclosure provides a battery testing method, comprising: Acquire shallow charge and discharge performance test data of the target battery, wherein the shallow charge and discharge performance test data includes test data corresponding to each preset shallow charge and discharge performance test item of the target battery; Based on the shallow charge and discharge performance test data, the full charge and discharge performance of the target battery is predicted by the trained battery test model to obtain the prediction data of the full charge and discharge performance test items of the target battery; wherein, the trained battery test model is trained using any one of the methods in the first aspect.

[0013] In a third aspect, the present disclosure provides a model training device, comprising: A first acquisition module is used to acquire electrical performance test data for a plurality of batteries, wherein the electrical performance test data of each battery includes shallow charge and discharge performance test data and full charge and discharge performance test data of the battery; a second acquisition module, which obtains a plurality of training samples using electrical performance test data based on each battery and a preset test item set, wherein the test item set includes a plurality of shallow charge and discharge performance test items and a full charge and discharge performance test item, and each training sample includes a first test value of a battery corresponding to at least one shallow charge and discharge performance test item and a second test value of the battery corresponding to the full charge and discharge performance test item; The training module is used to continuously perform training operations on the battery performance test model based on multiple training samples until a preset training end condition is met to obtain a trained battery performance test model; wherein the training operation includes: For each training sample, based on the first test values ​​in the training sample, the full charge and discharge performance of the battery corresponding to the training sample is predicted by the battery performance test model to obtain the predicted data of the full charge and discharge performance test item; based on the difference between the second test value corresponding to each training sample and the predicted data, the training loss is determined, and the model parameters of the battery performance test model are adjusted based on the training loss.

[0014] In a fourth aspect, the present disclosure provides a battery testing device, comprising: A third acquisition module is used to acquire shallow charge and discharge performance test data of the target battery, wherein the shallow charge and discharge performance test data includes test data corresponding to each preset shallow charge and discharge performance test item of the target battery; A prediction module is used to predict the full charge and discharge performance of the target battery based on the shallow charge and discharge performance test data through a trained battery test model, and obtain prediction data of the full charge and discharge performance test items of the target battery; wherein the trained battery test model is trained using any method of the first aspect.

[0015] In a fifth aspect, the present disclosure provides an electronic device, comprising: a processor, a memory, and a bus; Bus, used to connect the processor and memory; a memory for storing operation instructions; The processor is configured to execute the method of the first aspect or the second aspect of the present disclosure by calling an operation instruction.

[0016] In a sixth aspect, the present disclosure provides a computer-readable storage medium storing a computer program, which is used to execute the method of the first aspect or the second aspect of the present disclosure.

[0017] In a seventh aspect, the present disclosure provides a computer program product, comprising a computer program, which implements the steps of the method of the first aspect or the second aspect of the present disclosure when the computer program is executed by a processor.

[0018] The technical solutions provided by the embodiments of the present disclosure have at least the following beneficial effects: Obtain electrical performance test data for multiple batteries; obtain multiple training samples based on the electrical performance test data of each battery and a preset set of test items; based on the multiple training samples, continuously perform training operations on the battery performance test model until a preset training end condition is met, thereby obtaining a trained battery performance test model; wherein the training operation includes: for each training sample, based on each first test value in the training sample, predicting the full charge and discharge performance of the battery corresponding to the training sample through the battery performance test model to obtain predicted data for the full charge and discharge performance test item; determining the training loss based on the difference between the second test value corresponding to each training sample and the predicted data, and adjusting the model parameters of the battery performance test model based on the training loss; in this way, in the application of the trained battery performance test model, the test results of the shallow charge and discharge performance of the target battery (shallow charge and discharge performance test data) are input into the trained battery performance test model to predict the test results of the full charge and discharge performance (predicted data of the full charge and discharge performance, such as the dynamic pressure difference at the end of discharge); thereby saving the test time of the battery for full charge and discharge performance testing and improving the test efficiency of the battery. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following briefly introduces the drawings required for describing the embodiments of the present disclosure.

[0020] Figure 1 A schematic diagram of the architecture of the model training system provided in an embodiment of the present disclosure; Figure 2 A flowchart of a model training method provided in an embodiment of the present disclosure; Figure 3 A schematic diagram of a perceptron network provided in an embodiment of the present disclosure; Figure 4 A schematic diagram of a training loss curve provided in an embodiment of the present disclosure; Figure 5 A schematic diagram showing the comparison between the predicted results and the actual results provided by the embodiment of the present disclosure; Figure 6 A flowchart of a battery testing method provided in an embodiment of the present disclosure; Figure 7 A flow chart of a method for predicting full charge and discharge performance provided by an embodiment of the present disclosure; Figure 8 A schematic diagram of full charge and discharge performance prediction provided by an embodiment of the present disclosure; Figure 9 A schematic diagram of the structure of a model training device provided in an embodiment of the present disclosure; Figure 10 A schematic structural diagram of a battery testing device provided in an embodiment of the present disclosure; Figure 11 A schematic structural diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] The following describes embodiments of the present disclosure in conjunction with the accompanying drawings. It should be understood that the embodiments described below in conjunction with the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present disclosure and do not constitute a limitation on the technical solutions of the embodiments of the present disclosure.

[0022] Those skilled in the art will understand that, unless otherwise stated, the singular forms "a", "an", "said", and "the" used herein may also include plural forms. It should be further understood that the terms "including" and "comprising" used in the embodiments of the present disclosure mean that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements, and / or components, but do not exclude implementation as other features, information, data, steps, operations, elements, components, and / or combinations thereof supported by the present technical field. It should be understood that when we say that an element is "connected" or "coupled" to another element, the element can be directly connected or coupled to the other element, or it can refer to the connection relationship between the element and the other element established through an intermediate element. In addition, the "connection" or "coupling" used here can include wireless connection or wireless coupling. The term "and / or" used here indicates at least one of the items defined by the term, for example, "A and / or B" indicates implementation as "A", or implementation as "B", or implementation as "A and B".

[0023] It is understandable that in the specific implementation of the present disclosure, data related to model training and battery testing is involved. When the above embodiments of the present disclosure are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0024] In order to make the objectives, technical solutions and advantages of the present disclosure more clear, the embodiments of the present disclosure will be further described in detail below with reference to the accompanying drawings.

[0025] The embodiment of the present disclosure is a model training method provided by a model training system, which relates to fields such as artificial intelligence.

[0026] In order to better understand and illustrate the solutions of the embodiments of the present disclosure, some technical terms involved in the embodiments of the present disclosure are briefly explained below.

[0027] Forward propagation algorithm: the training set Input model for training; training set The data feature X is used as the input layer to enter the model. After weighted summation and bias term b, the value of the hidden layer neuron is obtained through the ReLU activation function f(·) ,calculate Formula (1) and formula (2) are as follows: Formula (1) Formula (2) in, is the weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer, is the bias term of the jth neuron in the hidden layer; the subsequent calculations from hidden layer to hidden layer, and from hidden layer to output layer, are all converted through weighted calculation and activation function to obtain the final predicted output P.

[0028] Back propagation algorithm: Use mean square error as the loss function L to measure the error between the true output Y and the predicted output P. The formula (3) for calculating the loss function L is as follows: Formula (3) In which, Represents the result of the j-th position in the predicted output P, Indicates the real output Y The result of the j-th position is then used to calculate the gradient of the loss function with respect to the weights and bias terms of each layer. The weights are iteratively updated using the gradient descent method, where n represents the number of training samples.

[0029] Apply the chain rule to calculate the output layer weight corresponding to the j-th position of the loss function and bias The gradient formula (4) and formula (5) are as follows: Formula (4) Formula (5) Formulas (6) and (7) for iterative weight updates using the gradient descent method are as follows: Formula (6) Formula (7) in, Represents the learning rate, provided by the model construction.

[0030] Data normalization: Min-Max normalization is used to normalize the data set Zoom to [ , ]Data interval, usually , , its calculation formula (8) is as follows: Formula (8) RMSE: Root Mean Square Error (RMSE) is a measure of the difference between the predicted value and the actual value.

[0031] IQR: IQR (InterQuartile Range), also known as the interquartile range, is a method in descriptive statistics to determine the difference between the third quartile and the first quartile.

[0032] Based on this, the embodiments of the present disclosure provide a model training method, a battery testing method, an apparatus, a device, a computer-readable storage medium, and a program product. The specific technical solutions will be described below.

[0033] The solutions provided by the embodiments of the present disclosure involve artificial intelligence technology. The technical solutions of the present disclosure are described in detail below using specific embodiments. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. The embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0034] In order to better understand the solution provided by the embodiment of the present disclosure, the solution is described below in conjunction with a specific application scenario.

[0035] In one embodiment, Figure 1 The schematic diagram of the architecture of a model training system applicable to the embodiment of the present disclosure is shown in FIG. It can be understood that the model training method provided by the embodiment of the present disclosure can be applied to but not limited to the following applications: Figure 1 In the application scenario shown.

[0036] In this example, Figure 1 As shown, the architecture of the model training system in this example may include but is not limited to a server 10, a terminal 20, and a database 30. The server 10, the terminal 20, and the database 30 may interact with each other through a network 40.

[0037] The server 10 obtains electrical performance test data for multiple batteries, where the electrical performance test data of each battery includes shallow charge and discharge performance test data and full charge and discharge performance test data of the battery; the server 10 obtains multiple training samples based on the electrical performance test data of each battery and a preset test item set, where the test item set includes multiple shallow charge and discharge performance test items and one full charge and discharge performance test item, and each training sample includes a first test value of a battery corresponding to at least one shallow charge and discharge performance test item, and a second test value of the battery corresponding to the full charge and discharge performance test item; the server 10 continuously performs training operations on the battery performance test model based on the multiple training samples until a preset training end condition is met, thereby obtaining a trained battery performance test model; wherein the training operation includes: for each training sample, based on each first test value in the training sample, predicting the full charge and discharge performance of the battery corresponding to the training sample through the battery performance test model to obtain predicted data for the full charge and discharge performance test item; determining the training loss based on the difference between the second test value corresponding to each training sample and the predicted data, and adjusting the model parameters of the battery performance test model based on the training loss. The server 10 stores a plurality of training samples in the database 30 , and displays the model parameters of the battery performance test model and the model training status of the battery performance test model to the user through the terminal 20 .

[0038] It is understood that the above is only an example and is not limited to this embodiment.

[0039] Among them, terminals include but are not limited to smartphones (such as Android phones, iOS phones, etc.), mobile phone simulators, tablet computers, laptops, digital broadcast receivers, MIDs (Mobile Internet Devices), PDAs (Personal Digital Assistants), intelligent voice interaction devices, smart home appliances, and vehicle-mounted terminals.

[0040] A server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server or server cluster that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0041] The aforementioned networks may include, but are not limited to, wired networks and wireless networks. Wired networks include local area networks, metropolitan area networks, and wide area networks, and wireless networks include Bluetooth, Wi-Fi, and other wireless communication networks. The specific network type may be determined based on actual application scenarios and is not limited here.

[0042] See also Figure 2 , Figure 2 The flow chart of a model training method provided by an embodiment of the present disclosure is shown, wherein the method can be executed by any electronic device, such as a server, etc.; as an optional implementation, the method can be executed by a server. For the convenience of description, in the description of some optional embodiments below, the method will be executed by a server as an example of the execution subject. Figure 2 As shown, the model training method provided by the embodiment of the present disclosure includes the following steps: S201 , obtaining electrical performance test data for a plurality of batteries, wherein the electrical performance test data for each battery includes shallow charge and discharge performance test data and full charge and discharge performance test data of the battery.

[0043] Specifically, the electrical performance test data of multiple batteries are shown in Table 1: Table 1: Electrical performance test data of multiple batteries

[0044] Among them, 03XXX001, 03XXX002, ... 03XXX380 represent the package codes of 380 batteries (battery packs), the battery's shallow charge and discharge performance test data such as voltage, temperature, stage discharge capacity, stage discharge energy, stage discharge pressure difference, charging energy, charging capacity, etc., and the battery's full charge and discharge performance test data such as the dynamic pressure difference at the end of discharge.

[0045] S202, based on the electrical performance test data of each battery and a preset test item set, multiple training samples are obtained, wherein the test item set includes multiple shallow charge and discharge performance test items and one full charge and discharge performance test item, and each training sample includes a first test value of a battery corresponding to at least one shallow charge and discharge performance test item, and a second test value of the battery corresponding to the full charge and discharge performance test item.

[0046] Specifically, there are a variety of shallow charge and discharge performance test items such as voltage, temperature, stage discharge capacity, stage discharge energy, stage discharge pressure difference, charging energy, charging capacity, etc., and a full charge and discharge performance test item such as the dynamic pressure difference at the end of discharge.

[0047] For example, a training sample includes the voltage test value, temperature test value and discharge end dynamic pressure difference test value corresponding to the battery with package code 03XXX001, wherein the voltage test value is a first test value of a shallow charge and discharge performance test item, and the discharge end dynamic pressure difference test value is a second test value of a full charge and discharge performance test item.

[0048] S203, continuously performing a training operation on the battery performance test model based on the multiple training samples until a preset training end condition is met, thereby obtaining a trained battery performance test model; wherein the training operation includes: For each training sample, based on the first test values ​​in the training sample, the full charge and discharge performance of the battery corresponding to the training sample is predicted by the battery performance test model to obtain the predicted data of the full charge and discharge performance test item; based on the difference between the second test value corresponding to each training sample and the predicted data, the training loss is determined, and the model parameters of the battery performance test model are adjusted based on the training loss.

[0049] Specifically, the battery performance test model is as follows: Figure 3 The perceptron network shown in the figure is a feedforward neural network consisting of an input layer, one or more hidden layers, and an output layer. Each layer is composed of several neurons, and the neurons between two adjacent layers interact with each other through weighted connections. The perceptron network makes predictions through the forward propagation algorithm and adjusts the weights through the backpropagation algorithm combined with the optimizer to minimize the prediction error.

[0050] For example, Figure 4 The training loss curve of the battery performance test model during the training process is shown. The forward propagation and backpropagation steps are repeated to train the battery performance test model on the training set (multiple training samples). The optimal model parameters are retained until the value of the loss function (training loss) of the battery performance test model is less than or equal to the preset threshold, or the battery performance test model completes the training round, thereby obtaining a trained battery performance test model.

[0051] For example, Figure 5 As shown in the figure, the predicted results of the trained battery test model for the dynamic pressure difference at the end of discharge on the test set are compared with the actual results; among them, the full charge and discharge performance test item of the battery is the dynamic pressure difference at the end of discharge.

[0052] In an embodiment of the present disclosure, electrical performance test data for a plurality of batteries is obtained; based on the electrical performance test data of each battery and a preset set of test items, a plurality of training samples are obtained; based on the plurality of training samples, a training operation is continuously performed on a battery performance test model until a preset training end condition is met, thereby obtaining a trained battery performance test model; wherein the training operation includes: for each training sample, based on each first test value in the training sample, predicting the full charge and discharge performance of the battery corresponding to the training sample through the battery performance test model, and obtaining predicted data for the full charge and discharge performance test item; determining a training loss based on the difference between the second test value corresponding to each training sample and the predicted data, and adjusting the model parameters of the battery performance test model based on the training loss; thus, in the application of the trained battery performance test model, the test results of the shallow charge and discharge performance of the target battery (shallow charge and discharge performance test data) are input into the trained battery performance test model, and the test results of the full charge and discharge performance (predicted data of the full charge and discharge performance, such as the dynamic pressure difference at the end of discharge) are predicted; thereby saving the test time for the full charge and discharge performance test of the battery and improving the test efficiency of the battery.

[0053] In one embodiment, a preset model set is obtained, where the model set includes a plurality of battery performance test models; Based on multiple training samples, the battery performance test model is continuously trained until a preset training end condition is met, thereby obtaining a trained battery performance test model, including: Based on the multiple training samples, the multiple battery performance test models in the model set are continuously trained until a preset training end condition is met, thereby obtaining the multiple trained battery performance test models; The model performance of the multiple battery performance test models after training is tested respectively, and the model with the best performance among the multiple battery performance test models after training is determined as the trained battery performance test model.

[0054] Specifically, for example, based on the number of hidden layers, the number of neurons, and the optimizer learning rate parameter, a grid combination method is used to construct multiple perceptrons with different network structures, i.e., multiple battery performance test models. For example, grid combination is a method of combining parameters, and the final number of combinations is the product of the number of set parameters. The custom items for the hidden layers are [a1, a2, a3], where a1, a2, and a3 represent different numbers of hidden layers, respectively; the custom items for the number of neurons are [b1, b2, b3], where b1, b2, and b3 represent different numbers of neurons, respectively; and the custom items for the optimizer learning rate parameter are [c1, c2, c3], where c1, c2, and c3 represent different optimizer learning rate parameters, respectively. After grid combination, the number of parameter combinations is 3 × 3 × 3 = 27. In this way, 27 perceptrons with different network structures, i.e., 27 battery performance test models, can be constructed. For example, a battery performance test model has a1 number of hidden layers, b2 number of neurons, and c3 optimizer learning rate parameter.

[0055] The preset training end condition is, for example, that the value of the loss function (training loss) of the battery performance test model is less than or equal to a preset threshold; the preset training end condition is, for example, that the battery performance test model completes a training round.

[0056] In one embodiment, the plurality of training samples are divided into a training set and a test set; Based on multiple training samples, training operations are continuously performed on multiple battery performance test models in the model set, including: Based on multiple training samples in the training set, training operations are continuously performed on multiple battery performance test models respectively; Among them, the trained battery performance test model is determined by the following method: Based on multiple training samples in the test set and preset model performance test indicators, the model performance of multiple trained battery performance test models is tested respectively to obtain the index value of the model performance test indicator corresponding to each trained battery performance test model; Based on the indicator values ​​corresponding to the trained battery performance test models, the model with the best model performance among the trained battery performance test models is determined as the trained battery performance test model.

[0057] Specifically, for example, multiple training samples are constructed as a dataset , the dataset Normalize the data in the data.

[0058] For example, the dataset The training set is divided into the ratio of 7:1:2 , validation set With the test set ; Among them, the training set For model training, For model evaluation after training, Used to evaluate the generalization performance of the model; for matrix, for matrix, for Matrix, N is the sum of the number of features X (at least one shallow charge and discharge performance test item, such as voltage, temperature, etc.) and feature Y (full charge and discharge performance test item, such as dynamic pressure difference at the end of discharge). Represents the training set , validation set , test set The number of training samples in is the total number of training samples. The structure is as follows:

[0059] in, is data feature X, which represents shallow charge and discharge performance test data; is data feature Y, which represents full charge and discharge performance test data.

[0060] For example, the preset model performance test indicator is RMSE, and the optimal model screening is evaluated using the RMSE function. The RMSE corresponding to multiple trained battery performance test models is calculated by formula (9), and the trained battery performance test model corresponding to the smallest RMSE among the multiple RMSEs is determined as the model with the best model performance, that is, the trained battery performance test model; wherein, formula (9) is as follows: Formula (9) in, represents the predicted value, Represents the true value, N represents the number of training samples in the test set; the predicted value is, for example, the predicted data of the full charge and discharge performance test item, and the true value is, for example, the second test value of the full charge and discharge performance test item.

[0061] It should be noted that the training task of creating a battery performance test model includes: inputting a training set, a training task name, the number of hidden layers of the battery performance test model, the number of neurons in the hidden layer, the optimizer learning rate parameters, etc. After creating the training task of the battery performance test model, the training task of the battery performance test model can be started. During the battery manufacturing process, new battery data generated after shallow charge and discharge and full charge and discharge tests can be pushed to the system-specified task through the system interface, and retraining work can be performed on the optimal model, with continuous iteration to improve the versatility and accuracy of the model.

[0062] In one embodiment, for each battery, the electrical performance test data of the battery is obtained by: Acquiring multiple original electrical performance test data obtained by performing electrical performance tests on the battery within a preset time period; For each test item in the test item set, extracting a test value of the test item from multiple original electrical performance test data of the battery based on a preset test value extraction method for the test item; Based on the test value of each test item in the extracted test item set, the electrical performance test data of the battery is determined.

[0063] Specifically, for example, the original electrical performance test data is obtained based on the original electrical performance test data file generated by the equipment manufacturer. For example, the original electrical performance test data file can be represented by a matrix, such as a 100×3600 matrix, where 100 represents the 100 cells in a battery (battery pack) and 3600 represents 3600 time points in an hour. If the original electrical performance test data file corresponds to a shallow charge and discharge performance test, then the elements in the matrix represent the shallow charge and discharge performance test data of a single cell at a certain time point. If the original electrical performance test data file corresponds to a full charge and discharge performance test, then the elements in the matrix represent the full charge and discharge performance test data of a single cell at a certain time point. The shallow charge and discharge performance test and full charge and discharge performance test of the same battery correspond to different original electrical performance test data files.

[0064] For example, the test item set includes a variety of shallow charge and discharge performance test items and full charge and discharge performance test items. The various shallow charge and discharge performance test items include temperature, voltage, stage discharge capacity, stage discharge energy, stage discharge pressure difference, charging energy, charging capacity, etc.; the full charge and discharge performance test items include dynamic pressure difference at the end of discharge, etc.

[0065] In one embodiment, the multiple shallow charge and discharge performance test items include at least two of temperature, voltage, stage discharge capacity, stage discharge energy, stage discharge pressure difference, charging energy, and charging capacity; the full charge and discharge performance test item includes the dynamic pressure difference at the end of discharge; For each test item of temperature and voltage, the test value extraction method corresponding to the test item includes extracting the maximum value or minimum value of the test value of the test item in the original electrical performance test data for multiple times, or extracting the difference between the maximum value and the minimum value of the test value of the test item in the original electrical performance test data for multiple times; For each test item of stage discharge capacity, stage discharge energy, charging energy, and charging capacity, a method for extracting a test value corresponding to the test item includes extracting a cumulative value of the test value of the test item from multiple original electrical performance test data; For each test item in the stage discharge pressure difference and the discharge end dynamic pressure difference, the test value extraction method corresponding to the test item includes extracting the difference between the maximum and minimum test values ​​of the test item in multiple original electrical performance test data.

[0066] Specifically, the test value extraction method corresponding to the test item includes, for example, positioning selection, interval accumulation algorithm, interval maximum algorithm, interval minimum algorithm, interval range algorithm, etc.

[0067] For example, in positioning selection, the temperature of the battery (battery pack) at the last time point in the shallow charge and discharge performance test (original electrical performance test data file) is selected as the test value corresponding to the test item temperature. The temperature of the battery (battery pack) at the last time point can be the highest charging temperature.

[0068] The calculation formula (10) of the interval accumulation algorithm is as follows: Formula (10) For example, there are five stages of shallow charge and discharge: 1C charge, 0.8C charge, 0.5C charge, 0.2C charge, and rest; the battery capacity of each single battery at the last time point of any of these five stages can be superimposed and the As the test value corresponding to the battery capacity of the test item; among them, represents the i-th single cell, and n represents the number of single cells included in the battery (battery pack).

[0069] The calculation formula (11) of the interval maximum algorithm is as follows: Formula (11) For example, there are five stages of shallow charge and discharge: 1C charge, 0.8C charge, 0.5C charge, 0.2C charge, and rest; the maximum temperature of each single battery at the last time point in any of these five stages can be taken as the test value corresponding to the test item temperature; where, represents the i-th single cell, and n represents the number of single cells included in the battery (battery pack).

[0070] The calculation formula (12) of the interval minimum algorithm is as follows: Formula (12) For example, there are five stages of shallow charge and discharge: 1C charge, 0.8C charge, 0.5C charge, 0.2C charge, and rest; the minimum temperature of each single battery at the last time point in any of these five stages can be taken as the test value corresponding to the test item temperature; where, represents the i-th single cell, and n represents the number of single cells included in the battery (battery pack).

[0071] The calculation formula (13) of the interval range algorithm is as follows: Formula (13) For example, the test value corresponding to the dynamic pressure difference at the end of the test item discharge is calculated using the interval range algorithm. For example, the test value corresponding to the stage discharge pressure difference of the test item is calculated using the interval range algorithm.

[0072] In one embodiment, for each battery, the electrical performance test data of the battery is obtained by: Acquiring electrical performance test data for the battery in each of a plurality of preset time periods; The electrical performance test data of the battery in the first time period among the multiple preset sections is used as the electrical performance test data of the battery.

[0073] Specifically, for example, the Pandas module in the Python programming language is used to load electrical performance test data for data preprocessing, which includes data deduplication, null value detection, and outlier processing. For example, data deduplication uses the battery (battery pack) package code as a deduplication identifier to extract all unique data entries.

[0074] In one embodiment, based on the electrical performance test data of each battery and a preset test item set, a plurality of training samples are obtained, including: For each battery, if it is determined based on the electrical performance test data and the test item set of the battery that the test value corresponding to any test item in the test item set of the battery is missing or the test value is abnormal, the electrical performance test data of the battery is deleted; A plurality of training samples are obtained based on the first test value of each retained battery corresponding to each shallow charge and discharge performance test item in the test item set and the second test value corresponding to the full charge and discharge performance test item.

[0075] Specifically, for example, in the case of null value detection, if the test value of any test item in the test item set corresponding to the battery with the package code of 03XXX001 has a null value or a value of 0, the electrical performance test data of the battery will be deleted from Table 1; wherein, the null value or the value of 0 indicates that the test value of the test item is missing.

[0076] For example, outlier processing uses the electrical performance test standard detection method and the IQR method to remove detected outliers; wherein, outliers indicate that the test value of the test item is abnormal.

[0077] For example, the standard detection method for electrical performance testing includes: for each electrical performance test feature (test item, such as voltage), deleting all electrical performance test data that is not within the normal value range of the electrical performance test feature.

[0078] For example, the IQR method includes: for a certain electrical performance test feature (test item, such as voltage), sorting the test values ​​of the electrical performance test feature in positive order (for example, sorting the 380 voltages for the test item voltage in Table 1 from small to large), naming the sorted samples sample1, sample2, ..., sample380, and then dividing the data into 4 groups according to quantity, that is, sample1-sample95 is group 1, sample96-sample190 is group 2, sample191-sample285 is group 3, and sample286-sample380 is group 4, where the value of sample95 is , the value of sample190 is , the value of sample285 is , and then the upper limit of normal value is obtained through calculation With lower limit , in the interval ( , ) are considered as abnormal values. The calculation process is shown in formula (14), formula (15) and formula (16): Formula (14) Formula (15) Formula (16) For example, the electrical performance test data of 380 batteries in Table 1 are preprocessed to obtain electrical performance test data of 327 batteries.

[0079] In one embodiment, a plurality of training samples are obtained based on the first test value of each battery corresponding to each shallow charge and discharge performance test item in the test item set and the second test value corresponding to the full charge and discharge performance test item, including: For each shallow charge and discharge performance test item among a plurality of shallow charge and discharge performance test items, determining a correlation between the shallow charge and discharge performance test item and a full charge and discharge performance test item based on first test data and second test data corresponding to the test item; wherein the first test data includes a first test value of each retained battery corresponding to the shallow charge and discharge performance test item, and the second test data includes a second test value of the full charge and discharge performance test item; For each shallow charge and discharge performance test item, if the correlation corresponding to the shallow charge and discharge performance test item is less than a preset correlation threshold, the shallow charge and discharge performance test item is deleted from the test item set; For each retained battery, a training sample is obtained based on the battery corresponding to the first test data and the battery corresponding to the second test data.

[0080] Specifically, for example, for each shallow charge and discharge performance test item among a plurality of shallow charge and discharge performance test items, based on the first test data and the second test data corresponding to the test item, the correlation between the shallow charge and discharge performance test item and the full charge and discharge performance test item is determined by the Pearson correlation analysis algorithm; the calculation formula (17) of the Pearson correlation analysis algorithm is as follows: Formula (17) Among them, x and y represent two variables respectively; Represents the correlation between x and y, and its value range is (-1, 1). The closer the absolute value of the correlation is to 1, the better the correlation is, and the closer it is to 0, the worse the correlation is. 、 They represent the average values ​​of x and y respectively, n represents the number of samples, and i represents the sample number; n is 327 for example; x is a shallow charge and discharge performance test item, and y is a full charge and discharge performance test item.

[0081] For example, for each shallow charge and discharge performance test item, if the correlation corresponding to the shallow charge and discharge performance test item is Less than the correlation threshold , then delete the shallow charge and discharge performance test item from the test item set; if the correlation corresponding to the shallow charge and discharge performance test item is Greater than or equal to the correlation threshold ,Right now , then retain the shallow charge and discharge performance test item in the test item set.

[0082] The application of the embodiments of the present disclosure has at least the following beneficial effects: (1) Intelligent data processing: Data preprocessing can quickly and accurately process raw data samples (raw electrical performance test data) from a data perspective, clean the raw data samples, and improve the quality of training data.

[0083] (2) Multi-dimensional data mining prediction: Introducing a multi-layer perceptron model (trained battery test model). Through deep feature extraction and pattern recognition, the multi-layer perceptron model can reveal the implicit correlation between data, accurately capture the subtle differences in battery performance changes, and conduct a comprehensive analysis of multi-dimensional data such as voltage, stage capacity, and pressure difference during the electrical measurement process, so as to more comprehensively evaluate and predict the electrical performance of lithium batteries.

[0084] (3) Autonomous model definition and training: A simple and fast method for building a multi-layer perceptron model is designed. The number of hidden layers, the number of neurons in each layer, and the learning rate can be customized according to needs. A selection list of the above three parameters can be constructed, and parameter combination and model construction can be performed through grid combination.

[0085] (4) Efficient data learning and model iteration: A one-stop data retraining solution is adopted. Model retraining behavior is triggered by simply adding original sample data, ensuring that it can quickly respond to changes in battery electrical measurement data, iterate the model in real time, and improve the versatility and long-term stability of the model.

[0086] See also Figure 6 , Figure 6 The flowchart of a battery testing method provided by an embodiment of the present disclosure is shown, wherein the method can be executed by any electronic device, such as a server, etc. As an optional implementation, the method can be executed by a server. For the convenience of description, in the description of some optional embodiments below, the method will be executed by a server as an example of the execution subject. Figure 6 As shown, the battery testing method provided by the embodiment of the present disclosure includes the following steps: S501 : Acquire shallow charge and discharge performance test data of a target battery, wherein the shallow charge and discharge performance test data includes test data corresponding to preset shallow charge and discharge performance test items of the target battery.

[0087] S502, based on the shallow charge and discharge performance test data, predict the full charge and discharge performance of the target battery through the trained battery test model to obtain the prediction data of the full charge and discharge performance test items of the target battery; wherein the trained battery test model is trained using any one of the methods in the first aspect.

[0088] Specifically, the full charge and discharge performance test items of the target battery are, for example, the dynamic voltage difference at the end of discharge.

[0089] The application of the embodiments of the present disclosure has at least the following beneficial effects: In the application of the trained battery performance test model, the shallow charge and discharge performance test results of the target battery (shallow charge and discharge performance test data) are input into the trained battery performance test model to predict the full charge and discharge performance test results (predicted data of full charge and discharge performance, such as the dynamic pressure difference at the end of discharge); thereby saving the test time of the battery for full charge and discharge performance testing and improving the battery testing efficiency.

[0090] In order to better understand the method provided by the embodiment of the present disclosure, the solution of the embodiment of the present disclosure is further described below with reference to examples of specific application scenarios.

[0091] In a specific application scenario embodiment, such as a battery full charge and discharge performance prediction scenario, see Figure 7 , shows a processing flow of a full charge and discharge performance prediction method, such as Figure 7 As shown, the processing flow of the full charge and discharge performance prediction method provided by the embodiment of the present disclosure includes the following steps: S701: The server obtains original electrical performance test data of multiple batteries.

[0092] Specifically, for example, Figure 8 As shown, the original data of shallow charge and discharge test and the original data of full charge and discharge test are obtained, that is, the original electrical performance test data of multiple batteries are obtained.

[0093] S702: The server determines the electrical performance test data of the plurality of batteries based on the original electrical performance test data of the plurality of batteries by using a preset test value extraction method of the test items.

[0094] Specifically, for example, Figure 8 As shown, the electrical test sample data feature extraction is to determine the electrical performance test data of multiple batteries as shown in Table 1.

[0095] S703: The server determines the retained electrical performance test data of each battery through data preprocessing based on the electrical performance test data of the multiple batteries.

[0096] Specifically, for example, Figure 8 As shown in the figure, data preprocessing; data preprocessing includes data deduplication, null value detection, outlier processing, etc.

[0097] S704: The server determines a plurality of training samples based on the retained electrical performance test data of each battery.

[0098] Specifically, for example, Figure 8As shown, sample data feature selection, that is: for each shallow charge and discharge performance test item in a plurality of shallow charge and discharge performance test items, based on the first test data and the second test data corresponding to the test item, determine the correlation between the shallow charge and discharge performance test item and the full charge and discharge performance test item; wherein the first test data includes the first test value of each retained battery corresponding to the shallow charge and discharge performance test item, and the second test data includes the second test value of the full charge and discharge performance test item; for each shallow charge and discharge performance test item, if the correlation corresponding to the shallow charge and discharge performance test item is less than a preset correlation threshold, then delete the shallow charge and discharge performance test item from the test item set; for each retained battery, a training sample is obtained based on the first test data and the second test data corresponding to the battery.

[0099] S705, the server continuously performs training operations on multiple battery performance test models in the model set based on multiple training samples until a preset training end condition is met, thereby obtaining multiple trained battery performance test models.

[0100] Specifically, for example, Figure 8 As shown, based on the number of hidden layers, the number of neurons and the learning rate (optimizer learning rate parameter), multiple perceptrons (multi-layer perceptrons) with different network structures are constructed through grid combination, that is, multiple battery performance test models; multiple battery performance test models are trained and evaluated to obtain multiple battery performance test models after training.

[0101] For example, grid combination is a method of combining parameters with each other, and the final number of combinations is the product of the number of set parameters; the custom items of the hidden layer are [a1, a2, a3], where a1, a2, and a3 represent different numbers of hidden layers respectively; the custom items of the number of neurons are [b1, b2, b3], where b1, b2, and b3 represent different numbers of neurons respectively; the custom items of the optimizer learning rate parameters are [c1, c2, c3], where c1, c2, and c3 represent different optimizer learning rate parameters respectively; the number of parameter combinations after grid combination is 3×3×3=27; in this way, 27 perceptrons with different network structures can be constructed, that is, 27 battery performance test models; for example, a battery performance test model has a1 number of hidden layers, b2 number of neurons, and c3 optimizer learning rate parameter.

[0102] S706, the server tests the model performance of the multiple trained battery performance test models respectively, and determines the model with the best performance among the multiple trained battery performance test models as the trained battery performance test model.

[0103] Specifically, for example, Figure 8As shown, the optimal model selection, that is, the preset model performance test index is RMSE, the optimal model screening is evaluated using the RMSE function, and the RMSE corresponding to multiple trained battery performance test models is calculated by formula (9), and the trained battery performance test model corresponding to the smallest RMSE among the multiple RMSEs is determined as the model with the best model performance, that is, the trained battery performance test model.

[0104] S707: The server obtains shallow charge and discharge performance test data of the target battery.

[0105] S708 , the server predicts the full charge and discharge performance of the target battery based on the shallow charge and discharge performance test data of the target battery using the trained battery test model, and obtains prediction data of the full charge and discharge performance test item of the target battery.

[0106] Specifically, for example, Figure 8 As shown, based on the shallow charge and discharge performance test data of the target battery, the trained battery test model is called through the model API (Application Programming Interface) to predict the full charge and discharge performance of the target battery, and the predicted data of the full charge and discharge performance test items of the target battery, that is, the prediction result, is obtained.

[0107] The application of the embodiments of the present disclosure has at least the following beneficial effects: In the application of the trained battery performance test model, the shallow charge and discharge performance test results of the target battery (shallow charge and discharge performance test data) are input into the trained battery performance test model to predict the full charge and discharge performance test results (predicted data of full charge and discharge performance, such as the dynamic pressure difference at the end of discharge); thereby saving the test time of the battery for full charge and discharge performance testing and improving the battery testing efficiency.

[0108] The present disclosure also provides a model training device. The structural diagram of the model training device is shown in FIG. Figure 9 As shown, the model training device 80 includes a first acquisition module 801, a second acquisition module 802 and a training module 803.

[0109] A first acquisition module 801 is configured to acquire electrical performance test data for a plurality of batteries, wherein the electrical performance test data for each battery includes shallow charge and discharge performance test data and full charge and discharge performance test data of the battery; A second acquisition module 802 is configured to obtain a plurality of training samples based on the electrical performance test data of each battery and a preset test item set, wherein the test item set includes a plurality of shallow charge and discharge performance test items and a full charge and discharge performance test item, and each training sample includes a first test value of a battery corresponding to at least one shallow charge and discharge performance test item and a second test value of the battery corresponding to the full charge and discharge performance test item; The training module 803 is configured to continuously perform training operations on the battery performance test model based on multiple training samples until a preset training end condition is met, thereby obtaining a trained battery performance test model. The training operations include: For each training sample, based on the first test values ​​in the training sample, the full charge and discharge performance of the battery corresponding to the training sample is predicted by the battery performance test model to obtain the predicted data of the full charge and discharge performance test item; based on the difference between the second test value corresponding to each training sample and the predicted data, the training loss is determined, and the model parameters of the battery performance test model are adjusted based on the training loss.

[0110] In one embodiment, the first acquisition module 801 is further configured to: acquire a preset model set, where the model set includes multiple battery performance test models; The training module 803 is specifically used to: Based on the multiple training samples, the multiple battery performance test models in the model set are continuously trained until a preset training end condition is met, thereby obtaining the multiple trained battery performance test models; The model performance of the multiple battery performance test models after training is tested respectively, and the model with the best performance among the multiple battery performance test models after training is determined as the trained battery performance test model.

[0111] In one embodiment, the training module 803 is specifically configured to: Divide multiple training samples into training sets and test sets; Based on multiple training samples in the training set, training operations are continuously performed on multiple battery performance test models respectively; Among them, the trained battery performance test model is determined by the following method: Based on multiple training samples in the test set and preset model performance test indicators, the model performance of multiple trained battery performance test models is tested respectively to obtain the index value of the model performance test indicator corresponding to each trained battery performance test model; Based on the indicator values ​​corresponding to the trained battery performance test models, the model with the best model performance among the trained battery performance test models is determined as the trained battery performance test model.

[0112] In one embodiment, for each battery, the electrical performance test data of the battery is obtained by: Acquiring multiple original electrical performance test data obtained by performing electrical performance tests on the battery within a preset time period; For each test item in the test item set, extracting a test value of the test item from multiple original electrical performance test data of the battery based on a preset test value extraction method for the test item; Based on the test value of each test item in the extracted test item set, the electrical performance test data of the battery is determined.

[0113] In one embodiment, the multiple shallow charge and discharge performance test items include at least two of temperature, voltage, stage discharge capacity, stage discharge energy, stage discharge pressure difference, charging energy, and charging capacity; the full charge and discharge performance test item includes the dynamic pressure difference at the end of discharge; For each test item of temperature and voltage, the test value extraction method corresponding to the test item includes extracting the maximum value or minimum value of the test value of the test item in the original electrical performance test data for multiple times, or extracting the difference between the maximum value and the minimum value of the test value of the test item in the original electrical performance test data for multiple times; For each test item of stage discharge capacity, stage discharge energy, charging energy, and charging capacity, a method for extracting a test value corresponding to the test item includes extracting a cumulative value of the test value of the test item from multiple original electrical performance test data; For each test item in the stage discharge pressure difference and the discharge end dynamic pressure difference, the test value extraction method corresponding to the test item includes extracting the difference between the maximum and minimum test values ​​of the test item in multiple original electrical performance test data.

[0114] In one embodiment, for each battery, the electrical performance test data of the battery is obtained by: Acquiring electrical performance test data for the battery in each of a plurality of preset time periods; The electrical performance test data of the battery in the first time period among the multiple preset sections is used as the electrical performance test data of the battery.

[0115] In one embodiment, the second acquisition module 802 is specifically configured to: For each battery, if it is determined based on the electrical performance test data and the test item set of the battery that the test value corresponding to any test item in the test item set of the battery is missing or the test value is abnormal, the electrical performance test data of the battery is deleted; A plurality of training samples are obtained based on the first test value of each retained battery corresponding to each shallow charge and discharge performance test item in the test item set and the second test value corresponding to the full charge and discharge performance test item.

[0116] In one embodiment, the second acquisition module 802 is specifically configured to: For each shallow charge and discharge performance test item among a plurality of shallow charge and discharge performance test items, determining a correlation between the shallow charge and discharge performance test item and a full charge and discharge performance test item based on first test data and second test data corresponding to the test item; wherein the first test data includes a first test value of each retained battery corresponding to the shallow charge and discharge performance test item, and the second test data includes a second test value of the full charge and discharge performance test item; For each shallow charge and discharge performance test item, if the correlation corresponding to the shallow charge and discharge performance test item is less than a preset correlation threshold, the shallow charge and discharge performance test item is deleted from the test item set; For each retained battery, a training sample is obtained based on the battery corresponding to the first test data and the battery corresponding to the second test data.

[0117] The application of the embodiments of the present disclosure has at least the following beneficial effects: Obtain electrical performance test data for multiple batteries; obtain multiple training samples based on the electrical performance test data of each battery and a preset set of test items; based on the multiple training samples, continuously perform training operations on the battery performance test model until a preset training end condition is met, thereby obtaining a trained battery performance test model; wherein the training operation includes: for each training sample, based on each first test value in the training sample, predicting the full charge and discharge performance of the battery corresponding to the training sample through the battery performance test model to obtain predicted data for the full charge and discharge performance test item; determining the training loss based on the difference between the second test value corresponding to each training sample and the predicted data, and adjusting the model parameters of the battery performance test model based on the training loss; in this way, in the application of the trained battery performance test model, the test results of the shallow charge and discharge performance of the target battery (shallow charge and discharge performance test data) are input into the trained battery performance test model to predict the test results of the full charge and discharge performance (predicted data of the full charge and discharge performance, such as the dynamic pressure difference at the end of discharge); thereby saving the test time of the battery for full charge and discharge performance testing and improving the test efficiency of the battery.

[0118] The present disclosure also provides a battery testing device. The structural diagram of the battery testing device is shown in FIG. Figure 10 As shown, the battery testing device 90 includes a third acquisition module 901 and a prediction module 902 .

[0119] A third acquisition module 901 is configured to acquire shallow charge and discharge performance test data of a target battery, wherein the shallow charge and discharge performance test data includes test data corresponding to each preset shallow charge and discharge performance test item of the target battery; Prediction module 902 is used to predict the full charge and discharge performance of the target battery based on the shallow charge and discharge performance test data through a trained battery test model, and obtain prediction data of the full charge and discharge performance test items of the target battery; wherein the trained battery test model is trained using any one of the methods of the first aspect.

[0120] The application of the embodiments of the present disclosure has at least the following beneficial effects: In the application of the trained battery performance test model, the shallow charge and discharge performance test results of the target battery (shallow charge and discharge performance test data) are input into the trained battery performance test model to predict the full charge and discharge performance test results (predicted data of full charge and discharge performance, such as the dynamic pressure difference at the end of discharge); thereby saving the test time of the battery for full charge and discharge performance testing and improving the battery testing efficiency.

[0121] The present disclosure also provides an electronic device. The structural diagram of the electronic device is as follows: Figure 11 As shown, Figure 11 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, which may be used for data exchange between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the number of transceivers 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present disclosure.

[0122] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the present disclosure. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, or a combination of a DSP and a microprocessor.

[0123] Bus 4002 may include a path for transmitting information between the above components. Bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. Bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0124] The memory 4003 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, without limitation herein.

[0125] The memory 4003 is used to store the computer program for executing the embodiments of the present disclosure, and the execution is controlled by the processor 4001. The processor 4001 is used to execute the computer program stored in the memory 4003 to implement the steps shown in the above method embodiments.

[0126] Among them, electronic equipment includes but is not limited to: servers, etc.

[0127] The application of the embodiments of the present disclosure has at least the following beneficial effects: Obtain electrical performance test data for multiple batteries; obtain multiple training samples based on the electrical performance test data of each battery and a preset set of test items; based on the multiple training samples, continuously perform training operations on the battery performance test model until a preset training end condition is met, thereby obtaining a trained battery performance test model; wherein the training operation includes: for each training sample, based on each first test value in the training sample, predicting the full charge and discharge performance of the battery corresponding to the training sample through the battery performance test model to obtain predicted data for the full charge and discharge performance test item; determining the training loss based on the difference between the second test value corresponding to each training sample and the predicted data, and adjusting the model parameters of the battery performance test model based on the training loss; in this way, in the application of the trained battery performance test model, the test results of the shallow charge and discharge performance of the target battery (shallow charge and discharge performance test data) are input into the trained battery performance test model to predict the test results of the full charge and discharge performance (predicted data of the full charge and discharge performance, such as the dynamic pressure difference at the end of discharge); thereby saving the test time of the battery for full charge and discharge performance testing and improving the test efficiency of the battery.

[0128] An embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment can be implemented.

[0129] The embodiments of the present disclosure further provide a computer program product, including a computer program, which can implement the steps and corresponding contents of the aforementioned method embodiments when executed by a processor.

[0130] It should be understood that, although the flowcharts of the embodiments of the present disclosure indicate the various operation steps by arrows, the order of implementation of these steps is not limited to the order indicated by the arrows. Unless otherwise clearly stated herein, in some implementation scenarios of the embodiments of the present disclosure, the implementation steps in each flowchart can be performed in other orders as required. In addition, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages based on the actual implementation scenario. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage in these sub-steps or stages can also be executed at different times. In scenarios where the execution times are different, the order of execution of these sub-steps or stages can be flexibly configured as required, and the embodiments of the present disclosure do not limit this.

[0131] The above description is only an optional implementation method for some implementation scenarios of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the solution of the present disclosure, other similar implementation methods based on the technical ideas of the present disclosure also fall within the protection scope of the embodiments of the present disclosure.

Claims

1. A model training method, characterized in that: include: Acquire electrical performance test data for a plurality of batteries, wherein the electrical performance test data for each battery includes shallow charge and discharge performance test data and full charge and discharge performance test data of the battery; Based on the electrical performance test data of each of the batteries and a preset test item set, a plurality of training samples are obtained, wherein the test item set includes a plurality of shallow charge and discharge performance test items and a full charge and discharge performance test item, and each of the training samples includes a first test value of a battery corresponding to at least one shallow charge and discharge performance test item, and a second test value of the battery corresponding to the full charge and discharge performance test item; Based on the multiple training samples, a training operation is continuously performed on the battery performance test model until a preset training end condition is met, thereby obtaining a trained battery performance test model; wherein the training operation includes: For each training sample, based on the first test values ​​in the training sample, the full charge and discharge performance of the battery corresponding to the training sample is predicted by the battery performance test model to obtain the predicted data of the full charge and discharge performance test item; based on the difference between the second test value corresponding to each training sample and the predicted data, the training loss is determined, and the model parameters of the battery performance test model are adjusted based on the training loss.

2. The method according to claim 1, characterized in that The method further comprises: Obtaining a preset model set, wherein the model set includes multiple battery performance test models; The method of continuously performing a training operation on the battery performance test model based on the multiple training samples until a preset training end condition is met to obtain a trained battery performance test model includes: Based on the multiple training samples, continuously perform training operations on the multiple battery performance test models in the model set until a preset training end condition is met, thereby obtaining multiple trained battery performance test models; The model performances of the multiple trained battery performance test models are tested respectively, and the model with the best performance among the multiple trained battery performance test models is determined as the trained battery performance test model.

3. The method according to claim 2, characterized in that The method further comprises: Dividing the plurality of training samples into a training set and a test set; The step of continuously performing training operations on the multiple battery performance test models in the model set based on the multiple training samples includes: Based on the multiple training samples in the training set, continuously perform training operations on multiple battery performance test models respectively; The trained battery performance test model is determined by: Based on the multiple training samples in the test set and the preset model performance test indicators, the model performance of the multiple trained battery performance test models is tested respectively to obtain the index value of the model performance test indicator corresponding to each trained battery performance test model; Based on the indicator values ​​corresponding to the trained battery performance test models, the model with the best model performance among the trained battery performance test models is determined as the trained battery performance test model.

4. The method according to claim 1, wherein For each of the batteries, the electrical performance test data of the battery is obtained by: Acquiring multiple original electrical performance test data obtained by performing electrical performance tests on the battery within a preset time period; For each test item in the test item set, extracting a test value of the test item from multiple original electrical performance test data of the battery based on a preset test value extraction method for the test item; Based on the extracted test value of each test item in the test item set, the electrical performance test data of the battery is determined.

5. The method according to claim 4, characterized in that The multiple shallow charge and discharge performance test items include at least two of temperature, voltage, stage discharge capacity, stage discharge energy, stage discharge pressure difference, charging energy, and charging capacity; the full charge and discharge performance test item includes the dynamic pressure difference at the discharge end; For each test item of temperature and voltage, the test value extraction method corresponding to the test item includes extracting the maximum value or minimum value of the test value of the test item in the original electrical performance test data for multiple times, or extracting the difference between the maximum value and the minimum value of the test value of the test item in the original electrical performance test data for multiple times; For each test item of stage discharge capacity, stage discharge energy, charging energy, and charging capacity, a method for extracting a test value corresponding to the test item includes extracting a cumulative value of the test value of the test item from multiple original electrical performance test data; For each test item in the stage discharge pressure difference and the discharge end dynamic pressure difference, the test value extraction method corresponding to the test item includes extracting the difference between the maximum and minimum test values ​​of the test item in multiple original electrical performance test data.

6. The method according to claim 1 or 4, characterized in that For each of the batteries, the electrical performance test data of the battery is obtained by: Acquiring electrical performance test data for the battery in each of a plurality of preset time periods; The electrical performance test data of the battery in the first time period among the plurality of preset sections is used as the electrical performance test data of the battery.

7. The method according to claim 1, characterized in that Based on the electrical performance test data of each battery and a preset test item set, a plurality of training samples are obtained, including: For each of the batteries, if it is determined, based on the electrical performance test data of the battery and the test item set, that a test value corresponding to any test item in the test item set for the battery is missing or abnormal, deleting the electrical performance test data of the battery; A plurality of training samples are obtained based on the first test value of each retained battery corresponding to each shallow charge and discharge performance test item in the test item set and the second test value corresponding to the full charge and discharge performance test item.

8. The method according to claim 7, characterized in that The obtaining of a plurality of training samples based on the first test value of each retained battery corresponding to each shallow charge and discharge performance test item in the test item set and the second test value corresponding to the full charge and discharge performance test item comprises: For each shallow charge and discharge performance test item among the multiple shallow charge and discharge performance test items, determining a correlation between the shallow charge and discharge performance test item and the full charge and discharge performance test item based on first test data and second test data corresponding to the test item; wherein the first test data includes a first test value of each retained battery corresponding to the shallow charge and discharge performance test item, and the second test data includes a second test value of the full charge and discharge performance test item; For each shallow charge and discharge performance test item, if the correlation corresponding to the shallow charge and discharge performance test item is less than a preset correlation threshold, deleting the shallow charge and discharge performance test item from the test item set; For each retained battery, a training sample is obtained based on the battery corresponding to the first test data and the battery corresponding to the second test data.

9. A battery testing method, characterized in that: include: Acquire shallow charge and discharge performance test data of a target battery, wherein the shallow charge and discharge performance test data includes test data corresponding to each preset shallow charge and discharge performance test item of the target battery; Based on the shallow charge and discharge performance test data, the full charge and discharge performance of the target battery is predicted by a trained battery test model to obtain prediction data of the full charge and discharge performance test items of the target battery; wherein the trained battery test model is trained using the method of any one of claims 1 to 8.

10. A model training device, characterized in that: include: A first acquisition module is used to acquire electrical performance test data for a plurality of batteries, wherein the electrical performance test data of each battery includes shallow charge and discharge performance test data and full charge and discharge performance test data of the battery; a second acquisition module, which obtains a plurality of training samples using electrical performance test data of each of the batteries and a preset test item set, wherein the test item set includes a plurality of shallow charge and discharge performance test items and a full charge and discharge performance test item, and each of the training samples includes a first test value of a battery corresponding to at least one shallow charge and discharge performance test item and a second test value of the battery corresponding to the full charge and discharge performance test item; A training module is configured to continuously perform a training operation on the battery performance test model based on the multiple training samples until a preset training end condition is met, thereby obtaining a trained battery performance test model; wherein the training operation includes: For each training sample, based on the first test values ​​in the training sample, the full charge and discharge performance of the battery corresponding to the training sample is predicted by the battery performance test model to obtain the predicted data of the full charge and discharge performance test item; based on the difference between the second test value corresponding to each training sample and the predicted data, the training loss is determined, and the model parameters of the battery performance test model are adjusted based on the training loss.

11. A battery testing device, characterized in that: include: a third acquisition module, configured to acquire shallow charge and discharge performance test data of a target battery, wherein the shallow charge and discharge performance test data includes test data corresponding to each preset shallow charge and discharge performance test item of the target battery; A prediction module is used to predict the full charge and discharge performance of the target battery based on the shallow charge and discharge performance test data through a trained battery test model to obtain prediction data for the full charge and discharge performance test item of the target battery; wherein the trained battery test model is trained using the method of any one of claims 1 to 8.

12. An electronic device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 9.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

14. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images