Battery fault diagnosis method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a battery pack fault diagnosis method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring the voltage and SOC value of a single battery in a to-be-detected battery pack as original data; extracting features of the original data by using wavelet packet transformation to obtain battery features; using a feature extraction unit in a pre-trained fault identification model to re-extract the battery features to obtain a battery extraction result; and inputting the battery extraction result into a fault identification unit in the fault identification model to obtain a battery fault result. By adopting the method, the fault type of the battery can be quickly and accurately identified in a model manner, and the identification efficiency and accuracy of the fault battery are improved.

Description

Technical Field

[0001] The present application relates to the technical field of electric vehicles, and in particular to a battery fault diagnosis method, device, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] As the power source of new energy vehicles, power batteries play a vital role. The performance of power batteries directly affects all aspects of the performance of the entire new energy vehicle. Due to the complex structure of the power battery system, it is usually composed of battery cells in series and parallel, and the overall performance of the battery pack is not a simple addition or amplification of the performance of the single cells. This situation is largely caused by the differences between the single cells, that is, the inconsistency problem of the battery. The inconsistency problems of the battery pack mainly include: small capacity, large internal resistance, and low SOC. As the degree of battery inconsistency increases, it is easy to induce lithium precipitation under long-term influence, increasing safety hazards such as internal short circuit and thermal runaway. Therefore, it is crucial to obtain relevant fault data for analysis and provide accurate fault diagnosis methods to maintain safe and stable operation of the battery.

[0003] At present, knowledge-based fault diagnosis methods rely on prior knowledge and require manual analysis, reasoning, judgment, etc. The diagnostic results of this method have low accuracy, low efficiency and strong subjectivity. Summary of the invention

[0004] Based on this, it is necessary to provide a battery pack fault diagnosis method, device, computer equipment, computer readable storage medium and computer program product that can accurately diagnose the battery fault type in order to solve the above technical problems.

[0005] In a first aspect, the present application provides a battery pack fault diagnosis method, comprising:

[0006] Obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data;

[0007] Extracting the features of the raw data using wavelet packet transform to obtain battery features;

[0008] Re-extracting the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0009] The battery extraction result is input into the fault identification unit in the fault identification model to obtain a battery fault result.

[0010] In one embodiment, the extracting the features of the raw data using wavelet packet transform to obtain battery features includes:

[0011] The original data is decomposed by three-layer wavelet packet transform to obtain three layers of low-frequency information and three layers of high-frequency information;

[0012] An energy characteristic value is determined based on the low-frequency information and the high-frequency information.

[0013] In one embodiment, the signal decomposition of the original data by using three-layer wavelet packet transform includes:

[0014] Decompose the original data to obtain the first layer of low-frequency information and the first layer of high-frequency information;

[0015] Decomposing the first layer of low-frequency information and the first layer of high-frequency information to obtain a second layer of low-frequency information and a second layer of high-frequency information;

[0016] The second layer of low-frequency information and the second layer of high-frequency information are decomposed to obtain a third layer of low-frequency information and a third layer of high-frequency information.

[0017] In one embodiment, the step of obtaining the voltage and SOC value of a single cell in the battery pack to be tested as raw data includes:

[0018] Carry out charge and discharge experiments on the battery to be tested through the battery testing unit;

[0019] The voltage and SOC value of the battery to be tested in the charge and discharge experiment are collected as raw data by a data collection unit.

[0020] In one embodiment, the training step of the fault identification model includes:

[0021] Connect the faulty battery and the normal battery to form a battery pack sample;

[0022] Charge and discharge the single cell samples in the battery pack samples to obtain the voltage and SOC value of the single cell samples as original sample data, extract the features of the original sample data by wavelet packet transform, and obtain battery feature samples; use the fault category of the single cell samples as labels;

[0023] The battery feature sample is input into the fault identification model to be trained to obtain a predicted fault category, a model loss is calculated based on the predicted fault category and the label, and parameters of the fault identification model to be trained are adjusted based on the model loss until the calculated model loss meets the training stop condition.

[0024] In one embodiment, the adjusting parameters of the fault identification model to be trained based on the model loss includes:

[0025] Based on the model loss, the Adam gradient descent method is used to adjust the parameters of the feature extraction module in the fault identification unit to be trained; and the particle swarm optimization algorithm is used to adjust the parameters of the fault identification module in the fault identification unit to be trained.

[0026] In a second aspect, the present application also provides a battery fault diagnosis device, comprising:

[0027] An acquisition module is used to acquire the voltage and SOC value of a single cell in the battery pack to be tested as raw data;

[0028] The primary extraction module uses wavelet packet transform to extract the features of the raw data to obtain battery features;

[0029] A re-extraction module, used to re-extract the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0030] The identification module is used to input the battery extraction result into the fault identification unit in the fault identification model to obtain the battery fault result.

[0031] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0032] Obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data;

[0033] Extracting the features of the raw data using wavelet packet transform to obtain battery features;

[0034] Re-extracting the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0035] The battery extraction result is input into the fault identification unit in the fault identification model to obtain a battery fault result.

[0036] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0037] Obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data;

[0038] Extracting the features of the raw data using wavelet packet transform to obtain battery features;

[0039] Re-extracting the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0040] The battery extraction result is input into the fault identification unit in the fault identification model to obtain a battery fault result.

[0041] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0042] Obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data;

[0043] Extracting the features of the raw data using wavelet packet transform to obtain battery features;

[0044] Re-extracting the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0045] The battery extraction result is input into the fault identification unit in the fault identification model to obtain a battery fault result.

[0046] The above-mentioned battery pack fault diagnosis method, device, computer equipment, computer-readable storage medium and computer program product first obtain the voltage and SOC value of the single battery in the battery pack to be tested as raw data; secondly, use wavelet packet transform to extract the features of the raw data to obtain battery features; then, use the feature extraction unit in the pre-trained fault recognition model to re-extract the battery features to obtain battery extraction results; finally, input the battery extraction results into the fault recognition unit in the fault recognition model to obtain the battery fault result. The fault category of the battery is quickly and accurately identified by the model, which improves the efficiency and accuracy of identifying faulty batteries. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 is a schematic flow chart of a battery pack fault diagnosis method in one embodiment;

[0049] Figure 2 is a structural diagram of a feature extraction unit in one embodiment;

[0050] Figure 3 is a schematic diagram of wavelet packet transform in one embodiment;

[0051] Figure 4 A schematic diagram of obtaining a battery pack voltage and SOC value in one embodiment;

[0052] Figure 5 A visualization diagram of features extracted by a feature extraction unit in one embodiment;

[0053] Figure 6 is a prediction result diagram of a fault identification model in one embodiment;

[0054] Figure 7 A schematic diagram of training and testing a fault identification model in one embodiment;

[0055] Figure 8 is a structural diagram of a battery pack fault diagnosis device in one embodiment;

[0056] Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] In one embodiment, Figure 1 As shown, a battery pack fault diagnosis method is provided. This embodiment takes the method applied to a terminal as an example for illustration. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0059] Step 102: Obtain the voltage and SOC value of the single battery in the battery pack to be tested as raw data.

[0060] Among them, SOC (State Of Charge) is the ratio of the available power in the battery to the nominal capacity, usually expressed as a percentage; its value range is 0~1. When SOC=0, it means that the battery is fully discharged, and when SOC=1, it means that the battery is fully charged. ‌

[0061] Exemplarily, different faulty batteries are connected in series with normal batteries, and a battery testing device and a data acquisition device are used to perform charge and discharge experiments on each battery to obtain the voltage and battery SOC value of all single cells as raw data.

[0062] Step 104: extract the features of the original data using wavelet packet transform to obtain battery features.

[0063] Among them, wavelet packet transform is a signal processing technology used to perform multi-scale refinement analysis of signals; it achieves detailed decomposition of the high-frequency and low-frequency parts of the signal by projecting the signal into the space spanned by the wavelet packet basis function.

[0064] Exemplarily, wavelet packet transform is used to decompose the original data, and signal energy is used to extract features of the signal to obtain battery features.

[0065] Step 106: Re-extract the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result.

[0066] Optional, such as Figure 2 As shown, the feature extraction unit can be a convolutional neural network. The size of the input layer is 16*1*1, there are 2 convolutional layers, 2 pooling layers, the size of the convolution kernel is 2*1, the number is 16, the size of the maximum pooling layer is 2*1, the step size is 1, and there is a fully connected layer of size 4. Relu is used as the activation function, the pooling operation adopts the maximum pooling operation mode, and a batch normalization layer is added between the convolution layer and the ReLU layer.

[0067] Exemplarily, the convolutional neural network in the pre-trained fault recognition model is used to re-extract the battery features to obtain the battery extraction results.

[0068] Step 108: input the battery extraction result into the fault identification unit in the fault identification model to obtain a battery fault result.

[0069] Optionally, the fault identification unit may be a support vector machine.

[0070] Exemplarily, the battery extraction result is input into a support vector machine in a pre-trained fault identification model to obtain a classification result of the battery fault.

[0071] Among them, the kernel function of the support vector machine adopts the radial basis function, and the formula is:

[0072]

[0073] in, and represents the input sample of the support vector machine, is the squared Euclidean distance between two samples, and g is a parameter of the radial basis function kernel.

[0074] The above-mentioned battery pack fault diagnosis method, device, computer equipment, computer-readable storage medium and computer program product first obtain the voltage and SOC value of the single battery in the battery pack to be tested as raw data; secondly, use wavelet packet transform to extract the features of the raw data to obtain battery features; then, use the feature extraction unit in the pre-trained fault recognition model to re-extract the battery features to obtain battery extraction results; finally, input the battery extraction results into the fault recognition unit in the fault recognition model to obtain the battery fault result. The fault category of the battery is quickly and accurately identified by the model, which improves the efficiency and accuracy of identifying faulty batteries.

[0075] In an exemplary embodiment, the extracting the features of the original data by wavelet packet transform to obtain the battery features includes:

[0076] The original data is decomposed by using three-layer wavelet packet transform to obtain three layers of low-frequency information and three layers of high-frequency information; based on the low-frequency information and the high-frequency information, the energy characteristic value is determined.

[0077] The original data are the voltage signal of the battery and the SOC value of the battery. The signal decomposition of the original data is to decompose the voltage signal of the battery and the SOC value of the battery respectively.

[0078] Exemplarily, the original data is decomposed using a three-layer wavelet packet transform to obtain three layers of low-frequency information and three layers of high-frequency information. The energy eigenvalue is determined using the wavelet packet coefficients of the third layer of low-frequency information and the third layer of high-frequency information, a total of 8 signals. The formula is as follows:

[0079]

[0080] in, The energy characteristic value of the jth signal in the i-th layer, is the jth signal of the i-th layer of wavelet packet decomposition, is the number of layers. In this embodiment, i is 3, and the signals of each node in the third layer are obtained. For the nodes, n is the The number of nodes of the layer wavelet packet, is the kth wavelet packet coefficient of the jth node.

[0081] In this embodiment, preliminary feature extraction is performed on metadata through wavelet packet transform, so that effective features can be extracted, which is beneficial to subsequent feature extraction and final classification.

[0082] In an exemplary embodiment, the signal decomposition of the original data by using three-layer wavelet packet transform includes:

[0083] The original data is decomposed to obtain a first layer of low-frequency information and a first layer of high-frequency information; the first layer of low-frequency information and the first layer of high-frequency information are decomposed to obtain a second layer of low-frequency information and a second layer of high-frequency information; the second layer of low-frequency information and the second layer of high-frequency information are decomposed to obtain a third layer of low-frequency information and a third layer of high-frequency information.

[0084] For example, Figure 3 As shown, the original data A(0,0) is decomposed to obtain the first layer of low-frequency information A(1,0) and the first layer of high-frequency information A(1,1); the first layer of low-frequency information and the first layer of high-frequency information are decomposed respectively to obtain the second layer of low-frequency information A(2,0) and the second layer of high-frequency information A(2,1), A(2,2) and A(2,3); the second layer of low-frequency information and high-frequency information are decomposed respectively to obtain the third layer of low-frequency information A(3,0) and the third layer of high-frequency information A(3,1), A(3,2), A(3,3), A(3,4), A(3,5), A(3,6) and A(3,7).

[0085] Among them, the low-frequency information A(3,0) of the third layer and the high-frequency information A(3,1), A(3,2), A(3,3), A(3,4), A(3,5), A(3,6) and A(3,7) of the third layer are used to determine the energy eigenvalues.

[0086] In this embodiment, eight wavelet packet signals of the third layer are obtained through three-layer decomposition, thereby obtaining a more refined signal representation and a more accurate feature extraction result.

[0087] In an exemplary embodiment, Figure 4 As shown, the step of obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data includes:

[0088] A charge and discharge experiment is performed on the battery to be tested by a battery testing unit; and a voltage and an SOC value of the battery to be tested in the charge and discharge experiment are collected as original data by a data acquisition unit.

[0089] For example, different faulty batteries are connected in series with normal batteries respectively; in the charge and discharge test, the voltage of all single batteries is collected by the battery test device and the data acquisition device, and the battery SOC value is obtained by the ampere-hour measurement method, and the voltage and SOC value are used as the original data. The formula for obtaining the SOC value is as follows:

[0090]

[0091] Among them, SOC(t) represents the SOC value of the battery at time t, represent The SOC value of the battery at the moment, Represents the rated capacity of the battery, represents the Coulomb efficiency, Representative from The integral value of the charge and discharge current I flowing through the battery circuit at time t.

[0092] In this embodiment, the voltage and SOC value of the battery are obtained as raw data to facilitate subsequent feature extraction of the data.

[0093] In an exemplary embodiment, the training step of the fault identification model includes:

[0094] Connect a faulty battery with a normal battery to form a battery pack sample; charge and discharge a single cell sample in the battery pack sample to obtain a voltage and SOC value of the single cell sample as original sample data, and use wavelet packet transform to extract features of the original sample data to obtain a battery feature sample; use the fault category of the single cell sample as a label; input the battery feature sample into a fault recognition model to be trained to obtain a predicted fault category, calculate a model loss based on the predicted fault category and the label, and adjust parameters of the fault recognition model to be trained based on the model loss until the calculated model loss meets the training stop condition.

[0095] Among them, the categories of single cells include: normal, large internal resistance, low SOC and small capacity.

[0096] Exemplarily, a faulty battery is connected to a normal battery to form a battery pack sample; the single cell samples in the battery pack sample are charged and discharged to obtain the voltage and SOC value of the single cell samples as original sample data, and the features of the original sample data are extracted using wavelet packet transform to obtain battery feature samples; the fault categories of the single cell samples are used as labels, including: normal, large internal resistance, low SOC and small capacity; the battery feature samples are input into a fault recognition model to be trained to obtain a predicted fault category, the model loss is calculated based on the predicted fault category and the label, and the parameters of the fault recognition model to be trained are adjusted based on the model loss until the calculated model loss meets the training stop condition.

[0097] In this embodiment, an ideal fault identification model can be obtained by training the fault identification model, so as to classify the faults of the battery to be tested.

[0098] In an exemplary embodiment, the step of adjusting parameters of the fault identification model to be trained based on the model loss includes:

[0099] Based on the model loss, the Adam gradient descent method is used to adjust the parameters of the feature extraction unit in the fault identification model to be trained; and the particle swarm optimization algorithm is used to adjust the parameters of the fault identification unit in the fault identification model to be trained.

[0100] Among them, the English name of the particle swarm optimization algorithm is Particle Swarm Optimization (PSO).

[0101] For example, based on the model loss, the Adam gradient descent method is used to adjust the parameters of the convolutional neural network in the fault recognition model to be trained. The parameter update formula is:

[0102]

[0103] Where: t is the number of training rounds, is the current parameter value, is the updated parameter value, is the step length, is the bias correction of the bias momentum of the gradient, is the bias correction of the second-order momentum of the gradient, Represents the stability parameter to prevent the denominator from being 0. The particle swarm optimization algorithm is used to adjust the parameters of the support vector machine in the fault recognition model to be trained. First, the random particles are initialized and the optimal solution is found through iteration. The particles update their own extreme values ​​and the global optimal solution by tracking the individual extreme value pbest and the global extreme value gbest. The formula for adjusting the speed and position is:

[0104]

[0105]

[0106] Where N is the total number of particles; is the particle speed; rand() is a random number between 0 and 1; xi is the current position of the particle; c1 and c2 are learning factors.

[0107] In this embodiment, the parameters of the fault identification model are updated by using the Adam and particle swarm acceleration algorithms, so that an ideal model can be obtained quickly for the classification of battery faults.

[0108] In an exemplary embodiment, Figure 5 As shown, after the features extracted by the feature extraction unit are reduced to three dimensions, a scatter plot can be drawn in the three-dimensional space, and the reduced data can be displayed in the three-dimensional space through the scatter plot. It can be seen that the feature data of different categories are clearly distinguishable, indicating that the feature extraction unit is effective in extracting the fault features of lithium-ion battery packs.

[0109] In an exemplary embodiment, Figure 6 As shown, when the prediction result of the horizontal axis is consistent with the true label of the sample of the test set of the vertical axis, the classification is correct. It can be seen that the prediction accuracy of normal batteries is 98.46%, the prediction accuracy of batteries with small capacity is 97.63%, the prediction accuracy of batteries with large internal resistance is 96.21%, and the prediction accuracy of batteries with low SOC is 99.11%.

[0110] In an exemplary embodiment, Figure 7 As shown, different faulty batteries are connected in series with normal batteries respectively; in the charge and discharge test, the voltage of all single batteries is collected by the battery test device and the data acquisition device, and the battery SOC value is obtained by the ampere-hour measurement method, and the voltage and SOC value are used as the original data. The formula for obtaining the SOC value is as follows:

[0111]

[0112] Among them, SOC(t) represents the SOC value of the battery at time t, represent The SOC value of the battery at the moment, Represents the rated capacity of the battery, represents the Coulomb efficiency, Representative from The integral value of the charge and discharge current I flowing through the battery circuit at time t. Decompose the original data A (0, 0) to obtain the first layer of low-frequency information A (1, 0) and the first layer of high-frequency information A (1, 1); decompose the first layer of low-frequency information and the first layer of high-frequency information respectively to obtain the second layer of low-frequency information A (2, 0) and the second layer of high-frequency information A (2, 1), A (2, 2) and A (2, 3); decompose the second layer of low-frequency information and high-frequency information respectively to obtain the third layer of low-frequency information A (3, 0) and the third layer of high-frequency information A (3, 1), A (3, 2), A (3, 3), A (3, 4), A (3, 5), A (3, 6) and A (3, 7). Use the wavelet packet coefficients of the third layer of low-frequency information and the third layer of high-frequency information, a total of 8 signals, to determine the energy eigenvalue, the formula is as follows:

[0113]

[0114] in, The energy characteristic value of the jth signal in the i-th layer, is the jth signal of the i-th layer of wavelet packet decomposition, is the number of layers. In this embodiment, i is 3, and the signals of each node in the third layer are obtained. For the nodes, n is the The number of nodes of the layer wavelet packet, is the kth wavelet packet coefficient of the jth node. The extracted battery features are divided into a test set and a training set in a ratio of 4:1. For the training set, the parameters of the feature extraction unit (Convolutional Neural Network, CNN) in the fault identification model (CNN-SVM fault diagnosis model) are first initialized, and then the battery features of the training set are passed through the convolution layer, pooling layer and fully connected layer, the loss is calculated, the parameters are updated using the Adam gradient descent method, and then the parameters of the SVM are optimized using the particle swarm optimization algorithm PSO to obtain the trained fault identification model. The convolutional neural network in the trained fault identification model is used to re-extract the battery features in the test set to obtain the battery extraction results. The battery extraction results are input into the support vector machine in the trained fault identification model to obtain the classification results of the battery fault.

[0115] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0116] In an exemplary embodiment, Figure 8 As shown, a battery pack fault diagnosis device is provided, comprising: an acquisition module, a primary extraction module, a secondary extraction module and an identification module, wherein:

[0117] An acquisition module is used to acquire the voltage and SOC value of a single cell in the battery pack to be tested as raw data;

[0118] A primary extraction module, used to extract the features of the original data using wavelet packet transform to obtain battery features;

[0119] A re-extraction module, used to re-extract the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0120] The identification module is used to input the battery extraction result into the fault identification unit in the fault identification model to obtain the battery fault result.

[0121] In one embodiment, the initial extraction module is further used to:

[0122] The original data is decomposed by using three-layer wavelet packet transform to obtain three layers of low-frequency information and three layers of high-frequency information; based on the low-frequency information and the high-frequency information, the energy characteristic value is determined.

[0123] In one embodiment, the initial extraction module is further used to:

[0124] The original data is decomposed to obtain a first layer of low-frequency information and a first layer of high-frequency information; the first layer of low-frequency information and the first layer of high-frequency information are decomposed to obtain a second layer of low-frequency information and a second layer of high-frequency information; the second layer of low-frequency information and the second layer of high-frequency information are decomposed to obtain a third layer of low-frequency information and a third layer of high-frequency information.

[0125] In one embodiment, the acquisition module is further used to:

[0126] A charge and discharge experiment is performed on the battery to be tested by a battery testing unit; and a voltage and an SOC value of the battery to be tested in the charge and discharge experiment are collected as original data by a data acquisition unit.

[0127] In one embodiment, it further includes: a training module for:

[0128] Connect a faulty battery with a normal battery to form a battery pack sample; charge and discharge a single cell sample in the battery pack sample to obtain a voltage and SOC value of the single cell sample as original sample data, and use wavelet packet transform to extract features of the original sample data to obtain a battery feature sample; use the fault category of the single cell sample as a label; input the battery feature sample into a fault recognition model to be trained to obtain a predicted fault category, calculate a model loss based on the predicted fault category and the label, and adjust parameters of the fault recognition model to be trained based on the model loss until the calculated model loss meets the training stop condition.

[0129] In one embodiment, the training module is further used to:

[0130] Based on the model loss, the Adam gradient descent method is used to adjust the parameters of the feature extraction unit in the fault identification model to be trained; and the particle swarm optimization algorithm is used to adjust the parameters of the fault identification unit in the fault identification model to be trained.

[0131] Each module in the above-mentioned battery pack fault diagnosis device can be implemented in whole or in part by software, hardware or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0132] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store raw data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a battery pack fault diagnosis method is implemented.

[0133] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0134] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0135] Obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data;

[0136] Extracting the features of the raw data using wavelet packet transform to obtain battery features;

[0137] Re-extracting the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0138] The battery extraction result is input into the fault identification unit in the fault identification model to obtain a battery fault result.

[0139] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0140] The original data is decomposed by using three-layer wavelet packet transform to obtain three layers of low-frequency information and three layers of high-frequency information; based on the low-frequency information and the high-frequency information, the energy characteristic value is determined.

[0141] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0142] The original data is decomposed to obtain a first layer of low-frequency information and a first layer of high-frequency information; the first layer of low-frequency information and the first layer of high-frequency information are decomposed to obtain a second layer of low-frequency information and a second layer of high-frequency information; the second layer of low-frequency information and the second layer of high-frequency information are decomposed to obtain a third layer of low-frequency information and a third layer of high-frequency information.

[0143] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0144] A charge and discharge experiment is performed on the battery to be tested by a battery testing unit; and a voltage and an SOC value of the battery to be tested in the charge and discharge experiment are collected as original data by a data acquisition unit.

[0145] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0146] Connect a faulty battery with a normal battery to form a battery pack sample; charge and discharge a single cell sample in the battery pack sample to obtain a voltage and SOC value of the single cell sample as original sample data, and use wavelet packet transform to extract features of the original sample data to obtain a battery feature sample; use the fault category of the single cell sample as a label; input the battery feature sample into a fault recognition model to be trained to obtain a predicted fault category, calculate a model loss based on the predicted fault category and the label, and adjust parameters of the fault recognition model to be trained based on the model loss until the calculated model loss meets the training stop condition.

[0147] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0148] Based on the model loss, the Adam gradient descent method is used to adjust the parameters of the feature extraction unit in the fault identification model to be trained; and the particle swarm optimization algorithm is used to adjust the parameters of the fault identification unit in the fault identification model to be trained.

[0149] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0150] Obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data;

[0151] Extracting the features of the raw data using wavelet packet transform to obtain battery features;

[0152] Re-extracting the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0153] The battery extraction result is input into the fault identification unit in the fault identification model to obtain a battery fault result.

[0154] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0155] The original data is decomposed by using three-layer wavelet packet transform to obtain three layers of low-frequency information and three layers of high-frequency information; based on the low-frequency information and the high-frequency information, the energy characteristic value is determined.

[0156] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0157] The original data is decomposed to obtain a first layer of low-frequency information and a first layer of high-frequency information; the first layer of low-frequency information and the first layer of high-frequency information are decomposed to obtain a second layer of low-frequency information and a second layer of high-frequency information; the second layer of low-frequency information and the second layer of high-frequency information are decomposed to obtain a third layer of low-frequency information and a third layer of high-frequency information.

[0158] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0159] A charge and discharge experiment is performed on the battery to be tested by a battery testing unit; and a voltage and an SOC value of the battery to be tested in the charge and discharge experiment are collected as original data by a data acquisition unit.

[0160] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0161] Connect a faulty battery with a normal battery to form a battery pack sample; charge and discharge a single cell sample in the battery pack sample to obtain a voltage and SOC value of the single cell sample as original sample data, and use wavelet packet transform to extract features of the original sample data to obtain a battery feature sample; use the fault category of the single cell sample as a label; input the battery feature sample into a fault recognition model to be trained to obtain a predicted fault category, calculate a model loss based on the predicted fault category and the label, and adjust parameters of the fault recognition model to be trained based on the model loss until the calculated model loss meets the training stop condition.

[0162] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0163] Based on the model loss, the Adam gradient descent method is used to adjust the parameters of the feature extraction unit in the fault identification model to be trained; and the particle swarm optimization algorithm is used to adjust the parameters of the fault identification unit in the fault identification model to be trained.

[0164] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0165] Obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data;

[0166] Extracting the features of the raw data using wavelet packet transform to obtain battery features;

[0167] Re-extracting the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result;

[0168] The battery extraction result is input into the fault identification unit in the fault identification model to obtain a battery fault result.

[0169] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0170] The original data is decomposed by using three-layer wavelet packet transform to obtain three layers of low-frequency information and three layers of high-frequency information; based on the low-frequency information and the high-frequency information, the energy characteristic value is determined.

[0171] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0172] The original data is decomposed to obtain a first layer of low-frequency information and a first layer of high-frequency information; the first layer of low-frequency information and the first layer of high-frequency information are decomposed to obtain a second layer of low-frequency information and a second layer of high-frequency information; the second layer of low-frequency information and the second layer of high-frequency information are decomposed to obtain a third layer of low-frequency information and a third layer of high-frequency information.

[0173] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0174] A charge and discharge experiment is performed on the battery to be tested by a battery testing unit; and a voltage and an SOC value of the battery to be tested in the charge and discharge experiment are collected as original data by a data acquisition unit.

[0175] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0176] Connect a faulty battery with a normal battery to form a battery pack sample; charge and discharge a single cell sample in the battery pack sample to obtain a voltage and SOC value of the single cell sample as original sample data, and use wavelet packet transform to extract features of the original sample data to obtain a battery feature sample; use the fault category of the single cell sample as a label; input the battery feature sample into a fault recognition model to be trained to obtain a predicted fault category, calculate a model loss based on the predicted fault category and the label, and adjust parameters of the fault recognition model to be trained based on the model loss until the calculated model loss meets the training stop condition.

[0177] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0178] Based on the model loss, the Adam gradient descent method is used to adjust the parameters of the feature extraction unit in the fault identification model to be trained; and the particle swarm optimization algorithm is used to adjust the parameters of the fault identification unit in the fault identification model to be trained.

[0179] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0180] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0181] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A battery pack fault diagnosis method, characterized in that: The method comprises: Obtaining the voltage and SOC value of the single battery in the battery pack to be tested as raw data; Extracting the features of the raw data using wavelet packet transform to obtain battery features; Re-extracting the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result; The battery extraction result is input into the fault identification unit in the fault identification model to obtain a battery fault result.

2. The method according to claim 1, characterized in that The method of extracting the features of the original data by wavelet packet transform to obtain battery features includes: The original data is decomposed by three-layer wavelet packet transform to obtain three layers of low-frequency information and three layers of high-frequency information; An energy characteristic value is determined based on the low-frequency information and the high-frequency information.

3. The method according to claim 2, characterized in that The method of performing signal decomposition on the original data by using three-layer wavelet packet transform comprises: Decompose the original data to obtain the first layer of low-frequency information and the first layer of high-frequency information; Decomposing the first layer of low-frequency information and the first layer of high-frequency information to obtain a second layer of low-frequency information and a second layer of high-frequency information; The second layer of low-frequency information and the second layer of high-frequency information are decomposed to obtain a third layer of low-frequency information and a third layer of high-frequency information.

4. The method according to claim 1, characterized in that: The step of obtaining the voltage and SOC value of a single cell in the battery pack to be tested as raw data includes: Carry out charge and discharge experiments on the battery to be tested through the battery testing unit; The voltage and SOC value of the battery to be tested in the charge and discharge experiment are collected as raw data by a data collection unit.

5. The method according to claim 1, characterized in that The training steps of the fault identification model include: Connect the faulty battery and the normal battery to form a battery pack sample; Charge and discharge the single cell samples in the battery pack samples to obtain the voltage and SOC value of the single cell samples as original sample data, extract the features of the original sample data by wavelet packet transform, and obtain battery feature samples; use the fault category of the single cell samples as labels; The battery feature sample is input into the fault identification model to be trained to obtain a predicted fault category, a model loss is calculated based on the predicted fault category and the label, and parameters of the fault identification model to be trained are adjusted based on the model loss until the calculated model loss meets the training stop condition.

6. The method according to claim 5, characterized in that The step of adjusting parameters of the fault identification model to be trained based on the model loss includes: Based on the model loss, the Adam gradient descent method is used to adjust the parameters of the feature extraction unit in the fault identification model to be trained; and the particle swarm optimization algorithm is used to adjust the parameters of the fault identification unit in the fault identification model to be trained.

7. A battery pack fault diagnosis device, characterized in that: The device comprises: An acquisition module is used to acquire the voltage and SOC value of a single cell in the battery pack to be tested as raw data; A primary extraction module, used to extract the features of the original data using wavelet packet transform to obtain battery features; A re-extraction module, used to re-extract the battery features using a feature extraction unit in a pre-trained fault recognition model to obtain a battery extraction result; The identification module is used to input the battery extraction result into the fault identification unit in the fault identification model to obtain the battery fault result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. 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 6 are implemented.

10. 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 6 are implemented.

Images