Power battery detection method and device and nonvolatile storage medium

Abstract

The invention discloses a power battery detection method and device and a nonvolatile storage medium. The method comprises the following steps: acquiring various operation data of a target power battery; multiple groups of text data used for describing multiple kinds of operation data are determined, and the multiple kinds of operation data are in one-to-one correspondence with the multiple groups of text data; the multiple sets of text data are imported into a target large language model, a detection result of the target power battery is output through the target large language model, the target large language model is obtained by conducting fine adjustment on parameters of the large language model through a fine adjustment database, and the fine adjustment database comprises sample text data and a sample detection result. The technical problem that the fault detection method of the power battery is low in efficiency is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of health monitoring and management of new energy power batteries, and in particular to a power battery detection method, device, and non-volatile storage medium. Background Art

[0002] In recent years, the world's attention to the environment and energy has been increasing, and electric vehicles have been widely recognized for their clean energy advantages. In particular, ternary lithium batteries have become the preferred energy storage device for electric vehicles due to their high energy density, excellent cycle performance and good low temperature tolerance. Despite this, electric vehicles frequently experience thermal runaway events due to battery system failures, which makes their safety a focus of attention. However, current battery maintenance mainly relies on reactive measures or regular inspections after a failure occurs, which usually requires a lot of expert experience and manual input, resulting in a lot of time and human resources consumed in the consultation, diagnosis, decision-making and service processes in the maintenance process.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] Embodiments of the present invention provide a power battery detection method, device, and non-volatile storage medium to at least solve the technical problem of low efficiency of a power battery fault detection method.

[0005] According to one aspect of an embodiment of the present invention, a power battery detection method is provided, comprising: obtaining multiple operating data of a target power battery; determining multiple sets of text data respectively used to describe the multiple operating data, wherein the multiple operating data correspond one-to-one to the multiple sets of text data; importing the multiple sets of text data into a target large language model, and the target large language model outputs the detection result of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of the large language model using a fine-tuning database, and the fine-tuning database includes sample text data and sample detection results. This embodiment uses the target large language model to detect the power battery, achieving the technical effect of improving detection efficiency.

[0006] Optionally, determining multiple groups of text data respectively used to describe multiple operating data includes: converting the multiple operating data into time series data respectively to obtain multiple time series data corresponding to the multiple operating data; extracting features from the multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data; converting the multiple feature data into text respectively to obtain multiple groups of text data, wherein the multiple groups of text data are used to describe the features of the multiple time series. This embodiment converts the operating data of the power battery into text data that can be recognized and analyzed by the target large language model, thereby improving the accuracy of the target large language model in fault detection.

[0007] Optionally, feature extraction is performed on multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data, including: extracting target feature data from target time series data in the multiple time series data in the following manner, and extracting features on multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data: decomposing the target time series data to obtain trend factors and seasonal factors of the target time series data, wherein the trend factor represents the changing trend of the target time series data, and the seasonal factor represents the periodic change of the target time series data; extracting features from the trend factor to obtain trend features of the target time series data; extracting features from the seasonal factor to obtain seasonal features of the target time series data; and obtaining target feature data based on the trend features of the target time series data and the seasonal features of the target time series data. This embodiment illustrates how to extract feature data of time series data to provide more accurate data for subsequent analysis.

[0008] Optionally, the target large language model is obtained by: obtaining sample text data and sample detection results; generating a fine-tuning database based on the sample text data and sample detection results; and fine-tuning the parameters of the large language model based on the fine-tuning database to obtain the target large language model. This embodiment illustrates how to obtain the target large language model based on the large language model, that is, to customize the large language model, better apply the large language model to the battery detection scenario, and achieve the technical effect of improving the detection accuracy.

[0009] Optionally, the parameters of the large language model are fine-tuned according to the fine-tuning database to obtain a target large language model, including: obtaining a first weight matrix corresponding to the parameters of the large language model; constructing a second weight matrix with the same dimension as the first weight matrix; adjusting the second weight matrix according to the fine-tuning database to determine the target large language model. This embodiment illustrates the process of fine-tuning the parameters of the large language model, which makes the large language model more suitable for power battery detection scenarios while ensuring that the knowledge learned by the original large language model is not lost, and achieves the technical effect of improving the detection accuracy of the target large language model.

[0010] Optionally, according to the fine-tuning database, the second weight matrix is ​​adjusted to determine the target large language model, including: based on a preset number of iterations, according to the fine-tuning database, iteratively training the large language model provided with the first weight matrix and the second weight matrix, and iteratively adjusting the second weight matrix to obtain a third weight matrix; according to the third weight matrix, determining the target large language model. This embodiment is a further adjustment of the parameters of the large language model, and the large language model obtained thereby can be better applied to the power battery detection scenario, achieving the technical effect of improving the power battery detection efficiency and the accuracy of the power battery detection results.

[0011] According to another aspect of an embodiment of the present invention, a power battery detection device is provided, including: a first acquisition module, used to acquire multiple operating data of a target power battery; a first determination module, used to determine multiple groups of text data respectively used to describe the multiple operating data, wherein the multiple operating data correspond one-to-one to the multiple groups of text data; a first import module, used to import the multiple groups of text data into a target large language model, and the target large language model outputs a detection result of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of a large language model by using a fine-tuning database, and the fine-tuning database includes sample text data and sample detection results.

[0012] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided. The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing any one of the power battery detection methods.

[0013] According to another aspect of an embodiment of the present invention, an electronic device is provided, comprising: one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the power battery detection methods.

[0014] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, any one of the above-mentioned power battery detection methods is implemented.

[0015] In an embodiment of the present invention, by acquiring multiple operating data of a target power battery; determining multiple sets of text data respectively used to describe the multiple operating data, wherein the multiple operating data correspond one to one with the multiple sets of text data; importing the multiple sets of text data into a target large language model, the target large language model outputs the detection result of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of the large language model by fine-tuning the database, and the fine-tuning database includes sample text data and sample detection results. The technical problem of low efficiency of the fault detection method of the power battery is solved, the purpose of quickly detecting possible faults of the power battery is achieved, and the technical result of improving the efficiency of power battery fault detection is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0017] Figure 1 is a flow chart of a power battery detection method provided according to an embodiment of the present invention;

[0018] Figure 2 is a schematic diagram of a power battery detection device provided according to an embodiment of the present invention;

[0019] Figure 3 is a schematic diagram of an electronic device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0021] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] For the convenience of description, some nouns or terms involved in the embodiments of the present application are explained below:

[0023] According to an embodiment of the present invention, a method embodiment of power battery detection is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0024] Figure 1 is a flow chart of a power battery detection method according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0025] Step S102, obtaining various operating data of the target power battery.

[0026] In this step, testing the power battery is actually analyzing various operating data of the power battery. When the operating data of the power battery deviates from the normal value, it is determined that the power battery is abnormal. When the power battery is running, it will generate various operating data, including but not limited to current, voltage, charge and discharge status, operating temperature and other data. These data are collected for subsequent testing steps.

[0027] Step S104 , determining a plurality of sets of text data respectively used to describe a plurality of types of operation data, wherein the plurality of types of operation data correspond one to one to the plurality of sets of text data.

[0028] In this step, the multiple operating data of the power battery contain multiple types of information, and the large language model is mainly used to model and process text data, and cannot directly process other types of information in the multiple operating data. Therefore, it is necessary to convert the multiple operating data of the power battery into multiple groups of text data that can be directly processed by the large language model, where each type of operating data corresponds to a group of text data.

[0029] In an optional embodiment, multiple groups of text data respectively used to describe multiple operating data are determined, including: converting the multiple operating data into time series data respectively, to obtain multiple time series data corresponding one-to-one to the multiple operating data; performing feature extraction on the multiple time series data respectively, to obtain multiple feature data corresponding one-to-one to the multiple time series data; converting the multiple feature data into text respectively, to obtain multiple groups of text data, wherein the multiple groups of text data are used to describe the respective features of the multiple time series.

[0030] Optionally, the multiple operating data of the power battery include but are not limited to current, voltage, temperature and other data, which change over time. Therefore, these data can be regarded as time series data.

[0031] For example, represent the current data as time series data:

[0032]

[0033] Among them, x 1 represents the current time series data vector, represents any value in the current time series data vector, n 1 Represents the length of the current time series data vector.

[0034] Represent voltage as time series data:

[0035] in x 2 represents the voltage time series data vector, Represents any data in the voltage time series data vector, n 2 Indicates the length of the voltage time series data vector.

[0036] Represent temperature as time series data;

[0037]

[0038] Among them, x 3 represents the temperature time series data vector, Represents any data in the temperature time series data vector, n 3 Represents the length of the temperature time series data vector.

[0039] According to the above method, all the operating data of the power battery are converted into corresponding time series data, and multiple time series data will be obtained. However, the multiple time series data obtained at this time cannot be recognized and analyzed by the large language model of power battery detection. It is necessary to extract feature information from each time series data and convert the feature information into text to obtain text data that can be recognized and analyzed by the large language model of power battery detection.

[0040] In an optional embodiment, feature extraction is performed on multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data one-to-one, including: extracting target feature data from target time series data in the multiple time series data in the following manner, and extracting feature on multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data one-to-one: decomposing the target time series data to obtain trend factors and seasonal factors of the target time series data, wherein the trend factor represents the changing trend of the target time series data, and the seasonal factor represents the periodic change of the target time series data; extracting feature of the trend factor to obtain trend feature of the target time series data; extracting feature of the seasonal factor to obtain seasonal feature of the target time series data; and obtaining target feature data based on the trend feature of the target time series data and the seasonal feature of the target time series data.

[0041] Optionally, data analysis can be performed on each time series to determine the overall trend of the distribution of data points in each time series data, that is, to determine the trend factor of the time series. At the same time, since the time series may be affected by temperature, humidity or other factors, this overall trend may contain multiple regular periodic changes. Therefore, after analyzing the overall trend of the distribution, the trend of periodic changes can also be analyzed, that is, to determine the seasonal factor of the time series. In addition, there may be some other situations that cause very few data in the time series to deviate from the complete trend and periodic trend. In other words, the trend factor is used to characterize the overall trend corresponding to the time series, the seasonal factor is used to characterize the regular periodic trend, and the residual factor is used to characterize the very few data that deviate from the complete trend and periodic trend.

[0042] By extracting features from the analyzed trend factors and seasonal factor distributions, trend features and seasonal features can be obtained, which can then be combined to obtain feature data for the corresponding time series data.

[0043] The time series calculation method is:

[0044] Specifically, for the sequence x k , k = 1, 2, 3, can be decomposed using the following formula

[0045] x k =t k +s k +r k

[0046] Among them, t k For the sequence x k The trend factor, s k For x kThe seasonal factor, r k For x k The residual factor of .

[0047] By extracting the features of trend factors and seasonal factors of time series, we can obtain the characteristic data of time series.

[0048] Step S106, importing multiple groups of text data into the target large language model, and the target large language model outputs the detection results of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of the large language model by using a fine-tuning database, and the fine-tuning database includes sample text data and sample detection results.

[0049] In this step, the sample detection result may correspond to the sample text data, and may be the result determined according to actual needs and output by the large language model after detecting the sample text data. The large language model may be a currently common open source large model. The sample text data is input into the open source large language model, and the detection result corresponding to the text data is input, that is, the sample text data and the sample detection result are used to fine-tune the parameters of the large language model. The open source large language model will learn based on the sample operation data and the sample detection result to obtain the power battery fault detection large language model required in this application. After obtaining the power battery fault detection large language model, the model can be used to detect the power battery. The detection process is to input the operation data of the power battery in the actual application scenario into the power battery fault detection large language model, and the power battery fault detection large language model will output the corresponding detection result.

[0050] In an optional embodiment, the target large language model is obtained by: acquiring sample text data and sample detection results; generating a fine-tuning database based on the sample text data and the sample detection results; and fine-tuning the parameters of the large language model based on the fine-tuning database to obtain the target large language model.

[0051] Optionally, the sample text data may be text used to describe the changing trend of the operating data of the sample battery, and the sample test results may be the annotation results of these text data, that is, the actual test results of the sample battery, which are used to guide how the open source large language model used in this application learns. The sample text data and the sample test results are stored in the fine-tuning database, and the open source large language model calls the sample text data and sample test results in the fine-tuning database to implement the training of the open source large language model and obtain the battery detection large language model required for this application. In addition, the fine-tuning database may also include a large amount of text data, including but not limited to books, academic literature, technical reports, expert knowledge and experience related to power battery health monitoring, health status assessment, and health maintenance, which are used to further train the language model.

[0052] In an optional embodiment, the parameters of the large language model are fine-tuned according to the fine-tuning database to obtain a target large language model, including: obtaining a first weight matrix corresponding to the parameters of the large language model; constructing a second weight matrix with the same dimension as the first weight matrix; and adjusting the second weight matrix according to the fine-tuning database to determine the target large language model.

[0053] Optionally, the parameters of the large language model include but are not limited to the structure of the model, the number of layers, the number of hidden units, the learning rate, etc. In the present application, the parameters of the large language model used for power battery detection can be the first weight matrix W. If the parameters in the first weight matrix are adjusted directly according to the fine-tuning database, it is possible that the already relatively complete large language model loses the previously learned content, thereby reducing the accuracy of the model output. Therefore, in order to make the large language model more suitable for the power battery detection scenario, another weight matrix can be introduced into the large language model, that is, an adapter is used to construct a second weight matrix A, and the dimension of the second weight matrix is ​​consistent with the first weight matrix W. When training using the fine-tuning database, only the second weight matrix can be adjusted to obtain a large language model that retains the knowledge in the original large language model and is also suitable for the power battery detection scenario.

[0054] In an optional embodiment, the second weight matrix is ​​adjusted according to a fine-tuning database to determine a target large language model, including: based on a preset number of iterations, according to the fine-tuning database, iteratively training a large language model provided with a first weight matrix and a second weight matrix, and iteratively adjusting the second weight matrix to obtain a third weight matrix; and determining the target large language model according to the third weight matrix.

[0055] Optionally, when training a large language model including a second weight matrix and a second weight matrix according to a fine-tuning database, a loss function can be set, and the second weight matrix can be iteratively adjusted according to the value of the loss function. When a pre-set iteration end condition is met, a trained target large language model can be obtained. The iteration end condition may include an iteration number condition and a loss value condition, wherein the iteration number condition means that when the iteration rounds reach a certain number, the iteration is stopped to obtain the final weight matrix; the loss value condition means that when a certain round of iterations ends, the loss value meets certain conditions, and the iteration is stopped to obtain the final weight matrix. This optional embodiment takes the iteration number condition as an example to perform iterative conditions on the second weight matrix. When determining the number of iterations, the number of samples in the fine-tuning database can be determined, and the number of iterative updates can be determined based on the number of samples. It should be noted that the third weight matrix refers to the weight matrix obtained by combining the first weight matrix and the second weight matrix.

[0056] The formula for iterative update is:

[0057] W FT =W+A

[0058] Among them, W FT Represents the third weight matrix obtained after fine-tuning. FT When , W always remains fixed, and A is updated with the number of iterations until convergence. The update formula of A is as follows:

[0059] A=U T V

[0060] Among them, U is the left decomposition matrix with row and column dimensions of m and d respectively, V is the right decomposition matrix with row and column dimensions of d and l respectively, and d is the rank of the adapter.

[0061] You can start the fine-tuning training program, set the training adapter rank d, fine-tuning iteration number, learning rate and other parameters; use the imported model-specific word segmenter to vectorize the fine-tuning database; start fine-tuning training until the preset loss value or fine-tuning iteration number is reached. End the fine-tuning training and save the final third weight matrix W FT , as parameters of the large language model for power battery detection.

[0062] Through the above, a variety of operating data of the target power battery is obtained; multiple groups of text data are determined to describe the various operating data, wherein the various operating data correspond to the multiple groups of text data one by one; the multiple groups of text data are imported into the target large language model, and the target large language model outputs the detection result of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of the large language model by fine-tuning the database, and the fine-tuning database includes sample text data and sample detection results. The technical problem of low efficiency of the fault detection method of the power battery is solved, and the purpose of quickly detecting possible faults of the power battery is achieved, thereby achieving the technical result of improving the efficiency of power battery fault detection.

[0063] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation mode, which may include the following steps:

[0064] Step S1, converting multiple operating data into time series data respectively, to obtain multiple time series data corresponding to the multiple operating data one by one; performing feature extraction on the multiple time series data respectively, to obtain multiple feature data corresponding to the multiple time series data one by one; converting the multiple feature data into text respectively, to obtain multiple groups of text data, wherein the multiple groups of text data are used to describe the respective features of the multiple time series.

[0065] Optionally, various operating data of the power battery include but are not limited to current, voltage, temperature and other data, and these data can be regarded as time series data.

[0066] For example, represent the current data as time series data:

[0067]

[0068] Among them, x 1 represents the current time series data vector, represents any value in the current time series data vector, n 1 Represents the length of the current time series data vector.

[0069] Represent voltage as time series data:

[0070] in x 2 represents the voltage time series data vector, Represents any data in the voltage time series data vector, n 2 Indicates the length of the voltage time series data vector.

[0071] Represent temperature as time series data;

[0072]

[0073] Among them, x 3 represents the temperature time series data vector, Represents any data in the temperature time series data vector, n 3 Represents the length of the temperature time series data vector.

[0074] According to the above method, all the operation data of the power battery are converted into corresponding time series data, and multiple time series data will be obtained. However, the multiple time series data obtained at this time cannot be recognized and analyzed by the large language model of power battery detection. It is necessary to extract feature information from each time series data to obtain text data that can be recognized and analyzed by the large language model of power battery detection.

[0075] Step S11, performing feature extraction on multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data one-to-one, including: extracting target feature data from target time series data in the multiple time series data in the following manner, performing feature extraction on multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data one-to-one: decomposing the target time series data to obtain trend factors and seasonal factors of the target time series data, wherein the trend factor represents the changing trend of the target time series data, and the seasonal factor represents the periodic change of the target time series data; performing feature extraction on the trend factor to obtain trend features of the target time series data; performing feature extraction on the seasonal factor to obtain seasonal features of the target time series data; obtaining target feature data according to the trend features of the target time series data and the seasonal features of the target time series data.

[0076] Optionally, data analysis can be performed on each time series to determine the overall trend of the distribution of data points in each time series data, which is determined as the trend factor of the time series. At the same time, since the time series may be affected by temperature, humidity or other factors, this overall trend may contain multiple regular periodic changes. Therefore, after analyzing the overall distribution trend of the data contained in the time series, the trend of periodic changes can also be analyzed, that is, the seasonal factor of the time series can be determined. In addition, there may be some other situations that cause very few data in the time series to deviate from the overall trend and periodic trend. In other words, the trend factor is used to characterize the overall trend corresponding to the time series, the seasonal factor is used to characterize the regular periodic trend, and the residual factor is used to characterize the very few data that deviate from the overall trend and periodic trend. The calculation method of the time series is:

[0077] For the sequence x k , k = 1, 2, 3, use the following formula to decompose:

[0078] x k =t k +s k +r k

[0079] Among them, t k For the sequence x k The trend factor, s k For x k The seasonal factor, r k For x k The residual factor of .

[0080] By extracting the features of trend factors and seasonal factors of time series, we can obtain the characteristic data of time series.

[0081] Step S2, obtaining sample text data and sample detection results; generating a fine-tuning database based on the sample text data and sample detection results; and fine-tuning the parameters of the large language model based on the fine-tuning database to obtain a target large language model.

[0082] Optionally, you can construct Query instructions, Answer instructions, and Instruction instructions, use these three instructions to build a template, obtain a fine-tuning database, train the large language model, and obtain a power battery detection large language model.

[0083] For example, {“Instruction”: “You are an expert in new energy power battery health monitoring and management. Please analyze based on the information given.”

[0084] "Query": "Analysis of the collected data shows that the frequency domain mean of the sequence is <c>, the peak value is <d>, the kurtosis value is <d>. What is the battery status at this time".

[0085] "Answer":"Battery status <status analysis result>Please continue to observe"}.

[0086] For example, {"Instruction":"You are an expert in new energy power battery health monitoring and management. Please analyze based on the information given."

[0087] "Query":"How should the battery <status> be maintained?"

[0088] "Answer":"<Maintenance plan>"}.

[0089] The above templates are not all the templates that have been constructed and need to be expanded and modified according to actual conditions to adapt to battery health management conditions.

[0090] Step S22, fine-tuning the parameters of the large language model according to the fine-tuning database to obtain the target large language model, including: obtaining a first weight matrix corresponding to the parameters of the large language model; constructing a second weight matrix with the same dimension as the first weight matrix; adjusting the second weight matrix according to the fine-tuning database to determine the target large language model.

[0091] Optionally, the parameters of the large language model include but are not limited to the model structure, number of layers, number of hidden units, learning rate, etc. The parameters of the large language model used for power battery detection in the present application can be counted as the first weight matrix W. If the parameters in the first weight matrix are adjusted directly according to the fine-tuning database, it may cause the relatively complete large prediction model to lose the previously learned content, thereby causing the model output accuracy to decrease. Another standard weight matrix is ​​introduced, that is, a second weight matrix A is constructed. The dimension of the second weight matrix is ​​consistent with the first weight matrix W. When the second weight matrix is ​​trained using the fine-tuning database, only the second weight matrix can be adjusted, so that a large language model that retains the original large language model and can be used for power battery detection scenarios can be obtained.

[0092] Step S23, adjusting the second weight matrix according to the fine-tuning database to determine the target large language model, including: based on a preset number of iterations, according to the fine-tuning database, iteratively training the large language model provided with the first weight matrix and the second weight matrix, and iteratively adjusting the second weight matrix to obtain a third weight matrix; determining the target large language model according to the third weight matrix.

[0093] Optionally, when training a large language model including a second weight matrix and a second weight matrix according to a fine-tuning database, a loss function can be set, and the second weight matrix can be iteratively adjusted according to the value of the loss function. When a pre-set iteration end condition is met, a trained target large language model can be obtained. The iteration end condition may include an iteration number condition and a loss value condition, wherein the iteration number condition means that when the iteration rounds reach a certain number, the iteration is stopped to obtain the final weight matrix; the loss value condition means that when a certain round of iterations ends, the loss value meets certain conditions, and the iteration is stopped to obtain the final weight matrix. This optional embodiment takes the iteration number condition as an example to perform iterative conditions on the second weight matrix. When determining the number of iterations, the number of samples in the fine-tuning database can be determined, and the number of iterative updates can be determined based on the number of samples. It should be noted that the third weight matrix refers to the weight matrix obtained by combining the first weight matrix and the second weight matrix.

[0094] The formula for iterative update is:

[0095] W FT =W+A,

[0096] Among them, W FT Represents the third weight matrix after fine-tuning. FT When , W always remains fixed, and A is updated with the number of iterations until convergence. The update formula of A is as follows:

[0097] A=U T V,

[0098] Among them, U is the left decomposition matrix with row and column dimensions of m and d respectively, V is the right decomposition matrix with row and column dimensions of d and l respectively, and d is the rank of the adapter.

[0099] You can start the fine-tuning training program, set the training adapter rank d, fine-tuning iteration number, learning rate and other parameters; use the imported model-specific word segmenter to vectorize the fine-tuning database; start fine-tuning training until the preset loss value or fine-tuning iteration number is reached. End the fine-tuning training and save the final third weight matrix W FT , as parameters of the large language model for power battery detection.

[0100] The above-mentioned optional implementation manner achieves at least the following effects: The present invention provides a power battery health detection method based on a large language model, which can translate the battery detection signal into a text description that can be understood by the large language model, and use the large language model that has been fine-tuned with professional data related to new energy power battery health detection and management for analysis, thereby achieving the technical effect of improving the efficiency of battery health detection.

[0101] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0102] In this embodiment, a power battery detection device is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated hereafter. As used below, the terms "module" and "device" may implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation in hardware, or a combination of software and hardware, is also possible and conceivable.

[0103] According to an embodiment of the present invention, a device embodiment for implementing a power battery detection method is also provided. Figure 2 is a schematic diagram of a power battery detection device according to an embodiment of the present invention. Figure 2 As shown, the power battery detection device includes a first acquisition module 21, a first determination module 22, and a first import module 23. The device is described below.

[0104] A first acquisition module 21, used to acquire various operating data of a target power battery;

[0105] A first determining module 22, connected to the first acquiring module, is used to determine a plurality of sets of text data respectively used to describe a plurality of operating data, wherein the plurality of operating data corresponds to the plurality of sets of text data in a one-to-one manner;

[0106] The first import module 23 is connected to the first determination module, and is used to import multiple groups of text data into a target large language model, and the target large language model outputs the detection result of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of the large language model by using a fine-tuning database, and the fine-tuning database includes sample text data and sample detection results.

[0107] In a power battery detection device provided by an embodiment of the present invention, a first acquisition module is set to obtain various operating data of a target power battery; a first determination module is used to determine multiple groups of text data respectively used to describe the various operating data, wherein the various operating data correspond one-to-one to the multiple groups of text data; a first import module is used to import the multiple groups of text data into a target large language model, and the target large language model outputs the detection result of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of the large language model by fine-tuning the database, and the fine-tuning database includes sample text data and sample detection results. The technical problem of low efficiency of the fault detection method of the power battery is solved, and the purpose of quickly detecting possible faults of the power battery is achieved, thereby achieving the technical result of improving the efficiency of power battery fault detection.

[0108] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0109] It should be noted that the first acquisition module 21, the first determination module 22, and the first import module 23 correspond to steps S102 to S106 in the embodiment, and the examples and application scenarios implemented by the modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments. It should be noted that the modules as part of the device can be run in a computer terminal.

[0110] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0111] The above-mentioned power battery detection device may also include a processor and a memory. The first acquisition module 21, the first determination module 22, the first import module 23, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to implement corresponding functions.

[0112] The processor includes a kernel, which retrieves the corresponding program unit from the memory. There can be one or more kernels. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip.

[0113] An embodiment of the present invention provides a non-volatile storage medium on which a program is stored. When the program is executed by a processor, a power battery detection method is implemented.

[0114] like Figure 3 As shown, an embodiment of the present invention provides an electronic device, the electronic device 10 includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: the memory is used to store a computer program, wherein when the computer program is executed by the processor, the processor implements the above-mentioned power battery detection method. The device in this article may be a server, a PC, etc.

[0115] The present invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that initializes the following method steps: computer instructions are executed by a processor to perform the above-mentioned power battery detection method.

[0116] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0118] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0120] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0121] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0122] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0123] It should also be noted that the term "includes", "comprising" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the existence of other identical elements in the process, method, commodity or device including the element.

[0124] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0125] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.< / d> < / d> < / c>

Claims

1. A power battery detection method, characterized in that: include: Obtain various operating data of the target power battery; Determining a plurality of sets of text data respectively used to describe the plurality of operating data, wherein the plurality of operating data corresponds one-to-one to the plurality of sets of text data; The multiple groups of text data are imported into a target large language model, and the target large language model outputs the detection result of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of the large language model by using a fine-tuning database, and the fine-tuning database includes sample text data and sample detection results.

2. The method according to claim 1, characterized in that The determining of a plurality of sets of text data respectively used to describe the plurality of operation data comprises: Converting the multiple types of operation data into time series data respectively, to obtain multiple time series data corresponding to the multiple types of operation data one by one; Extracting features from the multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data one by one; The plurality of feature data are respectively converted into texts to obtain the plurality of groups of text data, wherein the plurality of groups of text data are used to describe the respective features of the plurality of time series.

3. The method according to claim 2, characterized in that Feature extraction is performed on the multiple time series data respectively to obtain multiple feature data corresponding to the multiple time series data, including: The target feature data is extracted from the target time series data in the multiple time series data by adopting the following method, and the multiple time series data are respectively subjected to feature extraction to obtain multiple feature data corresponding to the multiple time series data one by one: Decomposing the target time series data to obtain a trend factor and a seasonal factor of the target time series data, wherein the trend factor represents a change trend of the target time series data, and the seasonal factor represents a periodic change of the target time series data; Extracting features of the trend factor to obtain trend features of the target time series data; Performing feature extraction on the seasonal factor to obtain seasonal features of the target time series data; The target characteristic data is obtained according to the trend characteristics of the target time series data and the seasonal characteristics of the target time series data.

4. The method according to claim 1, 2 or 3, characterized in that: The target large language model is obtained in the following way: Acquire the sample text data and the sample detection result; Generating the fine-tuning database according to the sample text data and the sample detection result; According to the fine-tuning database, the parameters of the large language model are fine-tuned to obtain the target large language model.

5. The method according to claim 4, characterized in that The step of fine-tuning the parameters of the large language model according to the fine-tuning database to obtain the target large language model includes: Obtaining a first weight matrix corresponding to the parameters of the large language model; Constructing a second weight matrix having the same dimension as the first weight matrix; The second weight matrix is ​​adjusted according to the fine-tuning database to determine the target large language model.

6. The method according to claim 5, characterized in that The step of adjusting the second weight matrix according to the fine-tuning database to determine the target large language model includes: Based on a preset number of iterations and according to the fine-tuning database, iteratively training a large language model provided with the first weight matrix and the second weight matrix, and iteratively adjusting the second weight matrix to obtain a third weight matrix; The target large language model is determined according to the third weight matrix.

7. A power battery detection device, characterized in that: include: A first acquisition module, used to acquire various operating data of a target power battery; A first determining module, used to determine a plurality of sets of text data respectively used to describe the plurality of operation data, wherein the plurality of operation data correspond to the plurality of sets of text data in a one-to-one manner; The first import module is used to import the multiple groups of text data into a target large language model, and the target large language model outputs the detection result of the target power battery, wherein the target large language model is obtained by fine-tuning the parameters of the large language model by using a fine-tuning database, and the fine-tuning database includes sample text data and sample detection results.

8. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the power battery detection method according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: One or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the power battery detection method described in any one of claims 1 to 6.

10. A computer program product comprising computer instructions, characterized in that: The computer instructions are executed by the processor to implement the power battery detection method described in any one of claims 1 to 6.

Images