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

Abstract

The invention relates to a battery fault identification method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring impedance spectrum data; the impedance spectrum data is obtained after preset excitation of a plurality of test frequency points is applied to the to-be-tested battery in a working state; converting the impedance spectrum data from a frequency domain to a time domain to obtain time domain impedance spectrum data; performing feature enhancement on the time domain impedance spectrum data to obtain enhanced impedance spectrum data; inputting the enhanced impedance spectrum data into a pre-trained fault recognition model to obtain a battery fault recognition result; the battery fault identification result is used for representing whether the to-be-detected battery has a fault or not. By adopting the method, whether the battery has obstacles can be accurately identified.

Description

Technical Field

[0001] The present application relates to the field of battery technology, and in particular to a battery fault identification method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the widespread application of lithium-ion batteries in new energy vehicles, energy storage power stations, portable devices, and other fields, the reliability and operational safety of battery systems have attracted widespread attention. Over long-term use, batteries may experience internal faults such as interface aging, SEI membrane damage, and lithium dendrite formation. Once these faults accumulate or develop, they can lead to serious consequences such as rapid capacity decay and thermal runaway. To proactively detect and accurately locate potential battery faults, the development of accurate, interpretable, and intelligent fault diagnosis methods is urgently needed.

[0003] Currently, traditional battery fault location methods primarily rely on voltage and current monitoring or simple impedance analysis. These methods typically only observe macroscopic performance changes, such as a drop in terminal voltage or an increase in internal resistance. However, due to the low resolution, high hysteresis, and susceptibility to temperature and load fluctuations of these signals, it is difficult to accurately identify the complex degradation mechanisms within the battery, leading to misjudgments. Summary of the Invention

[0004] Based on this, it is necessary to provide a battery fault identification method, device, computer equipment, computer-readable storage medium and computer program product that can accurately identify battery faults in order to address the above technical problems.

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

[0006] Obtaining impedance spectrum data; the impedance spectrum data is obtained by applying preset excitations at multiple test frequency points to the battery under test in a working state;

[0007] Converting the impedance spectrum data from the frequency domain to the time domain to obtain time domain impedance spectrum data;

[0008] Performing feature enhancement on the time-domain impedance spectrum data to obtain enhanced impedance spectrum data;

[0009] The enhanced impedance spectrum data is input into a pre-trained fault identification model to obtain a battery fault identification result; the battery fault identification result is used to characterize whether the battery to be tested has a fault.

[0010] In one embodiment, converting the impedance spectrum data from the frequency domain to the time domain to obtain the time domain impedance spectrum data includes:

[0011] Performing discrete modeling on the impedance spectrum data in combination with a relaxation time distribution model to construct a linear equation;

[0012] Solving the linear equation to obtain a relaxation time distribution vector;

[0013] Based on the relaxation time distribution vector, the impedance response characteristics on the time domain scale are inferred, and the time domain impedance spectrum data is obtained based on the impedance response characteristics.

[0014] In one embodiment, the feature enhancement of the time-domain impedance spectrum data to obtain enhanced impedance spectrum data includes:

[0015] Construct a preset direction vector;

[0016] The time-domain impedance spectrum data is projected onto the preset direction vector to obtain the enhanced impedance spectrum data.

[0017] In one embodiment, the above-mentioned construction of the preset direction vector includes:

[0018] Obtain impedance spectrum samples;

[0019] Calculating the distance between each impedance spectrum sample and the classification hyperplane, and adding the reciprocals of the distances to obtain the objective function;

[0020] Solve the objective function to obtain the preset direction vector.

[0021] In one embodiment, after obtaining the impedance spectrum data, the method further includes:

[0022] Preprocessing the impedance spectrum data; the preprocessing includes denoising, normalization and standardization;

[0023] The preprocessing of the impedance spectrum data is obtained by a pre-trained preprocessing model; the training process of the preprocessing model includes:

[0024] Obtaining first impedance spectrum training samples of multiple labeled state categories;

[0025] adding preset noise to each of the first impedance spectrum training samples;

[0026] Calculating the frequency attention weight of the first impedance spectrum training sample after adding random noise at each frequency point, and performing weighted processing on the impedance information of each frequency point according to the frequency attention weight to obtain a weighted impedance spectrum feature;

[0027] Inputting the weighted impedance spectrum feature into an initial encoder to extract a first feature vector;

[0028] The first feature vector is input into an initial decoder for impedance reconstruction to obtain a reconstruction result, and the parameters of the initial decoder and the initial encoder are adjusted according to the error between the first impedance spectrum training sample and the reconstruction result until the training is completed to obtain the preprocessing model.

[0029] In one embodiment, the training process of the fault identification model includes:

[0030] Obtaining a second impedance spectrum training sample; wherein the second impedance spectrum training sample carries a fault label;

[0031] Inputting the second impedance spectrum training sample into the initial model for training to obtain a fault prediction result;

[0032] A target loss function is obtained based on the predicted fault result and the fault label, and the parameters of the initial model are adjusted according to the target loss function until the training is completed to obtain the fault recognition model.

[0033] In a second aspect, the present application further provides a fault identification device, comprising:

[0034] An acquisition module is used to acquire impedance spectrum data; the impedance spectrum data is obtained by applying preset excitations at multiple test frequency points to the battery under test in a working state;

[0035] A conversion module, configured to convert the impedance spectrum data from the frequency domain to the time domain to obtain time domain impedance spectrum data;

[0036] A feature enhancement module, configured to enhance the features of the time-domain impedance spectrum data to obtain enhanced impedance spectrum data;

[0037] The identification module is used to input the enhanced impedance spectrum data into a pre-trained fault identification model to obtain a battery fault identification result; the battery fault identification result is used to indicate whether the battery to be tested has a fault.

[0038] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method in any one of the above embodiments when executing the computer program.

[0039] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method in any one of the above embodiments when the computer program is executed by a processor.

[0040] In a fifth aspect, the present application also provides a computer program product, comprising a computer program, which implements the steps of the method in any one of the above embodiments when executed by a processor.

[0041] The above-mentioned battery fault identification method, device, computer equipment, computer-readable storage medium and computer program product can quickly and accurately identify battery faults by first converting the frequency domain impedance spectrum into a time domain response that is more physically interpretable, then using features to enhance and amplify the signal dimensions that are highly correlated with the fault, and inputting them into the trained fault identification for identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. 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 any creative work.

[0043] Figure 1 1 is a flow chart of a battery fault identification method according to an embodiment;

[0044] Figure 2 is a flowchart of a battery fault identification method according to another embodiment;

[0045] Figure 3 is a structural block diagram of a battery fault identification device in one embodiment;

[0046] Figure 4 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0048] In one embodiment, Figure 1 As shown, a battery fault identification method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0049] Step 102 , obtaining impedance spectrum data. The impedance spectrum data is obtained by applying preset excitations at multiple test frequency points to the battery under test in a working state.

[0050] Impedance spectrum data refers to applying a sinusoidal excitation current or voltage across a range of test frequencies, such as a few millihertz to hundreds of kilohertz, to a battery system and recording the system's output voltage or current response. The complex impedance obtained at each test frequency is used as a data set to form an impedance spectrum.

[0051] Optionally, when the battery is in operation, a preset excitation is applied at multiple test frequency points, and then the phase difference and amplitude change of the voltage and current are tested to obtain the impedance value of the battery at each frequency.

[0052] The preset excitation refers to a defined AC disturbance signal applied to the battery under test during the impedance spectroscopy test. It is used to stimulate the battery's electrochemical response characteristics at different frequencies. It can be a small-amplitude sinusoidal voltage signal (such as 5-10mV), a constant-amplitude sinusoidal current signal, or other AC excitation signals that vary within a certain frequency range.

[0053] An impedance analyzer can be used to apply AC excitation to the lithium-ion battery under test. For example, under constant temperature conditions of 50% SOC and 25°C, a sinusoidal AC voltage signal with an amplitude of 10mV is applied to the battery. The frequency range is scanned from 100kHz to 10mHz in sequence, and the battery's response current is collected at each frequency point to calculate the complex impedance value at the corresponding frequency, and then obtain the impedance spectrum data.

[0054] Step 104: convert the impedance spectrum data from the frequency domain to the time domain to obtain time domain impedance spectrum data.

[0055] Impedance spectroscopy models the system's behavior in the frequency domain, but in real-world battery operation, the voltage and current in the time domain change over time. Therefore, to restore the true electrochemical behavior, impedance spectroscopy data must be converted from the frequency domain to the time domain, generating time-domain impedance spectroscopy data.

[0056] Among them, the time domain impedance spectrum data refers to a set of time domain signals reconstructed based on the frequency domain impedance spectrum, which describes the response characteristics of the battery to specific electrical excitation on the time axis.

[0057] Optionally, the impedance spectroscopy data can be converted from the frequency domain to the time domain via the relaxation time distribution.

[0058] Optionally, the complex impedance spectrum data in the frequency domain can be restored to a time-domain voltage or current response signal through Fourier transform or inverse Fourier transform, thereby obtaining the time-domain impedance spectrum data.

[0059] The impedance spectrum data is converted from the frequency domain to the time domain to obtain the time domain impedance spectrum data.

[0060] Alternatively, an equivalent circuit model can be fitted with the impedance spectrum data, and then the equivalent circuit model can be imported into circuit simulation software to simulate arbitrary excitation in the time domain to obtain time domain impedance spectrum data.

[0061] Alternatively, the transfer function constructed from the impedance spectrum data can be transformed from the frequency domain to the time domain using methods such as the Z transform or Laplace transform, thereby constructing a differential equation model for controller design or embedded simulation. For example, the impedance spectrum data can be fitted to a complex impedance function or RC network circuit model, and then processed using an inverse Laplace transform or differentiation to obtain the system's dynamic response function in the time domain, thereby forming the time-domain impedance spectrum data.

[0062] Step 106: Perform feature enhancement on the time-domain impedance spectrum data to obtain enhanced impedance spectrum data.

[0063] Since the time-domain impedance spectroscopy data includes a large amount of temporal variation details, the information that truly reflects the TV status may be submerged in high-dimensional redundancy or noise. Therefore, in this embodiment, the time-domain impedance spectroscopy data is feature-enhanced to obtain enhanced impedance spectroscopy data.

[0064] Alternatively, linear mapping can be performed on the time-domain impedance spectroscopy data, compressing the original high-dimensional time-domain response vector into a low-dimensional feature space through matrix transformation. For example, the time-domain impedance spectroscopy data can be mapped onto preset directional vectors. By calculating the projection values ​​of the original data onto these directions, the response changes associated with the state can be enhanced.

[0065] Optionally, based on the frequency-domain-time-domain joint feature extraction strategy, the frequency-domain impedance spectrum information is first embedded into the time-domain response, and then the joint time series is locally weighted or derivative-enhanced to amplify weak but critical response changes.

[0066] Step 108 : Input the enhanced impedance spectrum data into a pre-trained fault identification model to obtain a battery fault identification result; the battery fault identification result is used to characterize whether the battery to be tested has a fault.

[0067] Optionally, the enhanced impedance spectrum data is input into a pre-trained fault identification model to obtain a corresponding battery fault identification result, and the fault identification result can be used to determine whether there is a fault inside the battery to be tested.

[0068] The fault identification model is pre-trained, and its training process can refer to the description in the following embodiment.

[0069] In the above-mentioned battery fault identification method, by first converting the frequency domain impedance spectrum into a more physically interpretable time domain response, then using features to enhance and amplify the signal dimensions that are highly correlated with the fault, and inputting it into the trained fault recognition for identification, battery faults can be identified quickly and accurately.

[0070] In one embodiment, the above-mentioned conversion of impedance spectrum data from the frequency domain to the time domain to obtain time-domain impedance spectrum data includes: discrete modeling of the impedance spectrum data in combination with a relaxation time distribution model to construct a linear equation; solving the linear equation to obtain a relaxation time distribution vector; based on the relaxation time distribution vector, inversely deducing the impedance response characteristics on the time domain scale, and obtaining the time-domain impedance spectrum data based on the impedance response characteristics.

[0071] First, the impedance spectrum data is expressed using the DRT model, as shown in formula (1):

[0072] Formula (1)

[0073] Formula (1) indicates that the frequency domain of the battery corresponds to multiple different relaxation time constants The superposition of the responses, is a constant The weight of .

[0074] Next, formula (1) is discretized, and the discretized form is shown in formula (2).

[0075] Formula (2)

[0076] in, is the relaxation time distribution model to be determined; is the ohmic resistance; Frequency The complex impedance under is the discretized DRT point; N and M represent the number of frequency points and relaxation time points, respectively; is the weight of the logarithmic time interval.

[0077] Next, solve the linear equation. Specifically, express formula (2) in matrix form:

[0078] Formula (3)

[0079] Where Z is the complex impedance vector, A is the kernel matrix, is the relaxation time distribution vector.

[0080] from Perform feature extraction to obtain impedance response characteristics.

[0081] Optionally, the impedance response characteristics include one or more of peak position, peak value and peak area.

[0082] The peak position is expressed as ; Peak value is expressed as ; Peak area is expressed as .

[0083] Furthermore, the time domain impedance spectrum data is constructed based on the impedance response characteristics: .

[0084] In the above embodiment, by combining the relaxation time distribution model to discretely model the impedance spectrum data and extract the time domain response characteristics, the frequency domain data can be converted into structured features with clear physical meaning and time scale explanatory power.

[0085] In one embodiment, the feature enhancement of the time-domain impedance spectrum data to obtain enhanced impedance spectrum data includes: constructing a preset direction vector; and projecting the time-domain impedance spectrum data onto the preset direction vector to obtain enhanced impedance spectrum data.

[0086] Optionally, the construction of the preset direction vector may be based on a distance weighted discrimination (DWD) method. DWD can extract a direction vector from the impedance spectrum sample that can maximally distinguish samples of different categories for subsequent feature enhancement processing.

[0087] Furthermore, constructing a preset direction vector includes: obtaining impedance spectrum samples; calculating the distance between each impedance spectrum sample and the classification hyperplane, and adding the reciprocals of each distance to obtain a target function; solving the target function to obtain the preset direction vector.

[0088] Optionally, a plurality of labeled impedance spectrum samples are first obtained, and then a classification hyperplane is constructed. The goal is to find a direction vector that separates samples of different categories as much as possible in this direction. This direction is the normal vector of the hyperplane.

[0089] Furthermore, for each sample, its projected distance on the hyperplane can be calculated, and the inverse of each distance is used as the discriminant cost to construct an optimization objective function. The optimization objective function is to minimize the weighted sum of the sum of the inverse of the projected distances of all samples and the norm of the direction vector.

[0090] For example, the impedance spectrum sample is obtained and the objective function is obtained as shown in formula (4):

[0091] Formula (4)

[0092] in is the feature weight; For samples The weighted distance to the hyperplane; n represents the number of samples; Fault / normal label.

[0093] Optionally, the time domain impedance spectrum data is projected onto a preset direction vector, as shown in formula (5).

[0094] Formula (5)

[0095] in, Represents the projection result, that is, the enhanced impedance spectrum data.

[0096] In one embodiment, a trained support vector regression (SVR) model is used to predict the enhanced impedance spectroscopy data.

[0097] For example, the projected eigenvectors are fed into a support vector regression (SVR) model for modeling. A nonlinear regression model is constructed using a kernel function to minimize the structural risk and relaxation error:

[0098] Formula (6)

[0099] in, 、 are the upper and lower bound slack variables of the e-th sample respectively; b is the bias term; and the above formula needs to be constrained:

[0100] Formula (7)

[0101] in Map the kernel function to high-dimensional space; is the insensitive band width; C is the penalty parameter; this constraint ensures that the predicted value and the true value Within the deviation range. Finally, the function is predicted. , is the Lagrange multiplier corresponding to the above two inequality constraints, and K( , ) is the kernel function:

[0102] Formula (8)

[0103] Using kernel function , and finally get:

[0104] Formula (9)

[0105] By learning the correspondence between the feature vector and the battery fault location or status label, accurate prediction and identification of battery internal faults can be achieved.

[0106] In one embodiment, the training process of the above-mentioned fault identification model includes: obtaining a second impedance spectrum training sample; the second impedance spectrum training sample carries a fault label; inputting the second impedance spectrum training sample into the initial model for training to obtain a predicted fault result; obtaining a target loss function based on the predicted fault result and the fault label, and adjusting the parameters of the initial model according to the target loss function until the training is completed to obtain the fault identification model.

[0107] First, a second impedance spectrum training sample is obtained for training. The second impedance spectrum training sample can be a plurality of representative historical impedance response data, where each sample carries corresponding fault label information to indicate the battery state type at the time of acquisition, such as normal, overheated, polarization abnormality, or other defined fault states. The second impedance spectrum training sample can also be training samples obtained by testing the battery under different operating conditions. Different operating conditions refer to testing the battery at different SOCs.

[0108] Subsequently, the second impedance spectrum training samples are sequentially input into the initial model, which generates a predicted fault result for each sample. The initial model can be a binary classification model such as a support vector machine (SVM) or logistic regression. The predicted results can be expressed as classification labels for specific fault types, fault probability distributions, health scores, and so on. Furthermore, the predicted fault results output by the model are compared with the actual fault labels carried in the corresponding samples, and the error between the two is calculated. Based on this error, a target loss function is constructed. This target loss function can take the form of cross-entropy loss, mean squared error loss, or a weighted loss function, and can be specifically set based on the characteristics of the fault identification task. Next, based on the target loss function, the parameters of the initial model are iteratively adjusted using a backpropagation algorithm and a gradient optimization strategy to continuously narrow the gap between the predicted results and the actual labels.

[0109] The above training process can be executed repeatedly for multiple batches or multiple training rounds until the target loss function meets the convergence condition or the training accuracy reaches the preset threshold, and finally a fault recognition model with stable performance and strong generalization ability is obtained.

[0110] In the above embodiment, a fault identification model capable of accurately and automatically identifying battery faults is obtained through model training.

[0111] In one embodiment, after obtaining the impedance spectrum data, the method further includes: preprocessing the impedance spectrum data; the preprocessing includes denoising, normalization, and standardization; the preprocessing of the impedance spectrum data is obtained by a pre-trained preprocessing model; the training process of the preprocessing model includes: obtaining first impedance spectrum training samples of multiple labeled state categories; adding preset noise to each first impedance spectrum training sample; calculating the frequency attention weight of the first impedance spectrum training sample after adding random noise at each frequency point, and weighting the impedance information of each frequency point according to the frequency attention weight to obtain a weighted impedance spectrum feature; inputting the weighted impedance spectrum feature into an initial encoder to extract a first eigenvector; inputting the first eigenvector into an initial decoder for impedance reconstruction to obtain a reconstruction result, and adjusting the parameters of the initial decoder and the initial encoder according to the error between the first impedance spectrum training sample and the reconstruction result until the training is completed to obtain the preprocessing model.

[0112] In this embodiment, the impedance spectrum data is pre-processed using a pre-trained pre-processing model.

[0113] Optionally, noise of a preset amplitude is artificially added to each first impedance spectrum training sample to simulate random perturbations that may occur during the actual acquisition process, thereby enhancing the robustness of the model. Subsequently, based on the noisy training samples, frequency attention weights are calculated at each frequency point. The frequency attention weights are used to measure the importance of each frequency point for state discrimination. The frequency points in the original impedance spectrum are weighted according to the frequency attention weights, resulting in a weighted impedance spectrum feature that incorporates significant frequency response characteristics.

[0114] The weighted impedance spectrum features are then fed into the initially constructed encoder network to extract a low-dimensional first eigenvector, which is used to compress the core structural information in the impedance spectrum data. This first eigenvector is a low-dimensional eigenvector. This first eigenvector is then fed into the initial decoder to reconstruct the impedance spectrum and compare the error with the original training sample. Based on the reconstruction error between the reconstructed result and the original sample, the encoder and decoder parameters are jointly optimized and trained using a backpropagation mechanism until the reconstruction error meets the preset convergence criteria.

[0115] After the training is completed, the preprocessing model used in the data preprocessing stage is obtained to realize automatic denoising, normalization and standardization of the impedance spectrum data.

[0116] For example, the collected EIS data Z is subjected to a denoising process using a frequency attention mechanism-denoising autoencoder (DAE):

[0117] Formula (10)

[0118] Where N is the number of sampling points, and then the data is injected with noise:

[0119] Formula (11)

[0120] in, , split the real and imaginary parts and splice them:

[0121] Formula (12)

[0122] Then the attention score is calculated for each frequency point k:

[0123] Formula (13)

[0124] and is the frequency attention weight vector; are the trainable parameters of the attention layer; is the attention hidden layer dimension, and the attention weight is obtained by Softmax normalization:

[0125] Formula (14)

[0126] The features of each frequency point are weighted and mapped to the latent space:

[0127] Formula (15)

[0128] in is the set of trainable parameters of the encoder; is the weight matrix; is the bias vector; It is a symmetrical fully connected layer + linear activation. Reconstruct the complex number:

[0129] Formula (16)

[0130] Calculate the loss function of the data:

[0131] Formula (17)

[0132] Calculate the loss function for each piece of data and continuously optimize the parameters and , and finally obtain the optimal denoised, normalized and standardized data.

[0133] In one exemplary embodiment, in combination Figure 2 As shown, a battery fault identification method is provided.

[0134] During the model preparation phase, EIS (Electrochemical Impedance Spectroscopy Data), also known as the impedance spectroscopy data mentioned in the previous examples, is first generated. The data is then preprocessed and normalized, a step that can be handled using a preprocessing model. Next, DRT inversion and peak feature extraction are performed, along with feature vector construction and DWD mapping. For details, refer to the frequency-domain to time-domain conversion and feature enhancement sections in the previous examples.

[0135] Afterwards, the enhanced impedance spectrum data is input into the SVR model for training. The training process may refer to the training process of the fault identification model in the above embodiment.

[0136] Finally, during the usage phase, the sample to be tested is subjected to DWD mapping and then input into the trained SVR model, that is, the fault identification model, to obtain the predicted value and then the battery fault identification result.

[0137] In the above-mentioned embodiment, multi-scale physical process features can be extracted from electrochemical impedance spectroscopy (EIS) data at high resolution without assuming a circuit model. DRT inversion reveals the dynamic behavior of sub-processes such as charge transfer and ion diffusion within the battery. Combined with DWD feature projection and the SVR nonlinear regression model, accurate prediction and spatial localization of battery fault types and locations are achieved. This method improves the model's discriminative ability and generalization performance in small sample scenarios, providing stronger physical interpretability and real-time early warning capabilities, significantly enhancing the fault identification efficiency and operational safety of the battery system.

[0138] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed 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 portion of steps or stages in other steps.

[0139] Based on the same inventive concept, embodiments of the present application also provide a battery fault identification device for implementing the aforementioned battery fault identification method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more battery fault identification device embodiments provided below can be found in the above-described limitations of the battery fault identification method and will not be further elaborated here.

[0140] In an exemplary embodiment, Figure 3 As shown, a battery fault identification device is provided, including: an acquisition module 100, a conversion module 200, a feature enhancement module 300 and an identification module 400, wherein:

[0141] The acquisition module 100 is used to acquire impedance spectrum data; the impedance spectrum data is obtained by applying preset excitations at multiple test frequency points to the battery under test in a working state.

[0142] The conversion module 200 is used to convert the impedance spectrum data from the frequency domain to the time domain to obtain time domain impedance spectrum data.

[0143] The feature enhancement module 300 is used to enhance the features of the time-domain impedance spectrum data to obtain enhanced impedance spectrum data.

[0144] The identification module 400 is used to input the enhanced impedance spectrum data into a pre-trained fault identification model to obtain a battery fault identification result; the battery fault identification result is used to indicate whether the battery to be tested has a fault.

[0145] In one embodiment, the conversion module includes:

[0146] The equation building unit is used to discretely model the impedance spectrum data in combination with the relaxation time distribution model and construct a linear equation.

[0147] The solving unit is used to solve the linear equation and obtain the relaxation time distribution vector.

[0148] The time domain conversion unit is used to infer the impedance response characteristics on the time domain scale based on the relaxation time distribution vector, and obtain the time domain impedance spectrum data based on the impedance response characteristics.

[0149] In one embodiment, the feature enhancement module includes:

[0150] Vector construction unit, used to construct a preset direction vector.

[0151] The projection unit is used to project the time-domain impedance spectrum data onto a preset direction vector to obtain enhanced impedance spectrum data.

[0152] In one embodiment, the vector construction unit includes:

[0153] The first sample acquisition subunit is used to acquire an impedance spectrum sample.

[0154] The objective function construction subunit is used to calculate the distance between each impedance spectrum sample and the classification hyperplane, and to add the reciprocals of each distance to obtain the objective function.

[0155] The function solving subunit is used to solve the objective function and obtain the preset direction vector.

[0156] In one embodiment, the apparatus further comprises:

[0157] The preprocessing module is used to preprocess the impedance spectrum data; the preprocessing includes denoising, normalization and standardization.

[0158] In one embodiment, the pre-processing module further includes:

[0159] The second sample acquisition unit is configured to acquire first impedance spectrum training samples of a plurality of labeled state categories.

[0160] The noise adding unit is used to add preset noise to each first impedance spectrum training sample.

[0161] The weight calculation unit is used to calculate the frequency attention weight of the first impedance spectrum training sample after adding random noise at each frequency point, and perform weighted processing on the impedance information of each frequency point according to the frequency attention weight to obtain a weighted impedance spectrum feature.

[0162] The vector extraction unit is used to input the weighted impedance spectrum feature into the initial encoder to extract the first feature vector.

[0163] The model optimization unit is used to input the first eigenvector into the initial decoder for impedance reconstruction to obtain a reconstruction result, and adjust the parameters of the initial decoder and the initial encoder according to the error between the first impedance spectrum training sample and the reconstruction result until the training is completed to obtain a preprocessing model.

[0164] In one embodiment, the above-mentioned apparatus includes a fault identification model training module, and the fault identification model training module includes:

[0165] The third sample acquisition unit is used to acquire a second impedance spectrum training sample; the second impedance spectrum training sample carries a fault label.

[0166] The training unit is used to input the second impedance spectrum training sample into the initial model for training to obtain a fault prediction result.

[0167] The parameter optimization unit is used to obtain the target loss function based on the predicted fault results and fault labels, and adjust the parameters of the initial model according to the target loss function until the training is completed to obtain the fault recognition model.

[0168] Each module in the battery fault identification device described above can be implemented in whole or in part through software, hardware, or a combination thereof. Each module 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 the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0169] 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. Figure 4 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and 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 computer program in the non-volatile storage medium. The database of the computer device is used to store impedance spectrum 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 fault identification method is implemented.

[0170] Those skilled in the art will understand that Figure 4 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 shown in the figure, or combine certain components, or have a different component arrangement.

[0171] In an exemplary embodiment, a computer device is provided, comprising 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: obtaining impedance spectrum data; the impedance spectrum data is obtained by applying a preset excitation at multiple test frequency points to a battery under test in a working state; converting the impedance spectrum data from the frequency domain to the time domain to obtain time-domain impedance spectrum data; performing feature enhancement on the time-domain impedance spectrum data to obtain enhanced impedance spectrum data; inputting the enhanced impedance spectrum data into a pre-trained fault identification model to obtain a battery fault identification result; and the battery fault identification result is used to characterize whether the battery under test has a fault.

[0172] In one embodiment, when the processor executes the computer program, the following steps are further implemented: discrete modeling of the impedance spectrum data is performed in combination with the relaxation time distribution model to construct a linear equation; the linear equation is solved to obtain a relaxation time distribution vector; based on the relaxation time distribution vector, the impedance response characteristics on the time domain scale are inferred, and time domain impedance spectrum data is obtained based on the impedance response characteristics.

[0173] In one embodiment, when the processor executes the computer program, the processor further implements the following steps: constructing a preset direction vector; and projecting the time-domain impedance spectrum data onto the preset direction vector to obtain enhanced impedance spectrum data.

[0174] In one embodiment, when the processor executes the computer program, the following steps are further implemented: obtaining impedance spectrum samples; calculating the distance between each impedance spectrum sample and the classification hyperplane, and adding the reciprocals of each distance to obtain a target function; solving the target function to obtain a preset direction vector.

[0175] In one embodiment, when the processor executes the computer program, the following steps are further implemented: preprocessing the impedance spectrum data; the preprocessing includes denoising, normalization, and standardization; the preprocessing of the impedance spectrum data is obtained by a pre-trained preprocessing model; the training process of the preprocessing model includes: obtaining first impedance spectrum training samples of multiple labeled state categories; adding preset noise to each first impedance spectrum training sample; calculating the frequency attention weight of the first impedance spectrum training sample after adding random noise at each frequency point, and weighting the impedance information of each frequency point according to the frequency attention weight to obtain a weighted impedance spectrum feature; inputting the weighted impedance spectrum feature into an initial encoder to extract a first eigenvector; inputting the first eigenvector into an initial decoder for impedance reconstruction to obtain a reconstruction result, and adjusting the parameters of the initial decoder and the initial encoder according to the error between the first impedance spectrum training sample and the reconstruction result until the training is completed to obtain the preprocessing model.

[0176] In one embodiment, when the processor executes the computer program, the following steps are further implemented: obtaining a second impedance spectrum training sample; the second impedance spectrum training sample carries a fault label; inputting the second impedance spectrum training sample into the initial model for training to obtain a predicted fault result; obtaining a target loss function based on the predicted fault result and the fault label, and adjusting the parameters of the initial model according to the target loss function until the training is completed to obtain a fault identification model.

[0177] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: discrete modeling of impedance spectrum data is performed in combination with a relaxation time distribution model to construct a linear equation; the linear equation is solved to obtain a relaxation time distribution vector; based on the relaxation time distribution vector, the impedance response characteristics on the time domain scale are inferred, and time domain impedance spectrum data is obtained based on the impedance response characteristics.

[0178] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: constructing a preset direction vector; projecting the time domain impedance spectrum data onto the preset direction vector to obtain enhanced impedance spectrum data.

[0179] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: obtaining impedance spectrum samples; calculating the distance between each impedance spectrum sample and the classification hyperplane, and adding the reciprocals of each distance to obtain a target function; solving the target function to obtain a preset direction vector.

[0180] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: preprocessing the impedance spectrum data; the preprocessing includes denoising, normalization, and standardization; the preprocessing of the impedance spectrum data is obtained by a pre-trained preprocessing model; the training process of the preprocessing model includes: obtaining first impedance spectrum training samples of multiple labeled state categories; adding preset noise to each first impedance spectrum training sample; calculating the frequency attention weight of the first impedance spectrum training sample after adding random noise at each frequency point, and weighting the impedance information of each frequency point according to the frequency attention weight to obtain a weighted impedance spectrum feature; inputting the weighted impedance spectrum feature into an initial encoder to extract a first eigenvector; inputting the first eigenvector into an initial decoder for impedance reconstruction to obtain a reconstruction result, and adjusting the parameters of the initial decoder and the initial encoder according to the error between the first impedance spectrum training sample and the reconstruction result until the training is completed to obtain the preprocessing model.

[0181] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: obtaining a second impedance spectrum training sample; the second impedance spectrum training sample carries a fault label; inputting the second impedance spectrum training sample into the initial model for training to obtain a predicted fault result; obtaining a target loss function based on the predicted fault result and the fault label, and adjusting the parameters of the initial model according to the target loss function until the training is completed to obtain a fault identification model.

[0182] In one embodiment, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the following steps: discretely modeling impedance spectrum data in combination with a relaxation time distribution model to construct a linear equation; solving the linear equation to obtain a relaxation time distribution vector; and inferring an impedance response characteristic on a time domain scale based on the relaxation time distribution vector, and obtaining time domain impedance spectrum data based on the impedance response characteristic.

[0183] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: constructing a preset direction vector; projecting the time domain impedance spectrum data onto the preset direction vector to obtain enhanced impedance spectrum data.

[0184] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: obtaining impedance spectrum samples; calculating the distance between each impedance spectrum sample and the classification hyperplane, and adding the reciprocals of each distance to obtain a target function; solving the target function to obtain a preset direction vector.

[0185] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: preprocessing the impedance spectrum data; the preprocessing includes denoising, normalization, and standardization; the preprocessing of the impedance spectrum data is obtained by a pre-trained preprocessing model; the training process of the preprocessing model includes: obtaining first impedance spectrum training samples of multiple labeled state categories; adding preset noise to each first impedance spectrum training sample; calculating the frequency attention weight of the first impedance spectrum training sample after adding random noise at each frequency point, and weighting the impedance information of each frequency point according to the frequency attention weight to obtain a weighted impedance spectrum feature; inputting the weighted impedance spectrum feature into an initial encoder to extract a first eigenvector; inputting the first eigenvector into an initial decoder for impedance reconstruction to obtain a reconstruction result, and adjusting the parameters of the initial decoder and the initial encoder according to the error between the first impedance spectrum training sample and the reconstruction result until the training is completed to obtain the preprocessing model.

[0186] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: obtaining a second impedance spectrum training sample; the second impedance spectrum training sample carries a fault label; inputting the second impedance spectrum training sample into the initial model for training to obtain a predicted fault result; obtaining a target loss function based on the predicted fault result and the fault label, and adjusting the parameters of the initial model according to the target loss function until the training is completed to obtain a fault identification model.

[0187] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. 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. In particular, any reference to memory, database, or other media used in the embodiments provided in this 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. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0188] The technical features of the above embodiments can be combined arbitrarily. In order 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.

[0189] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A battery fault identification method, characterized in that: The method comprises: Obtaining impedance spectrum data; the impedance spectrum data is obtained by applying preset excitations at multiple test frequency points to the battery under test in a working state; Converting the impedance spectrum data from the frequency domain to the time domain to obtain time domain impedance spectrum data; Performing feature enhancement on the time-domain impedance spectrum data to obtain enhanced impedance spectrum data; The enhanced impedance spectrum data is input into a pre-trained fault identification model to obtain a battery fault identification result; the battery fault identification result is used to characterize whether the battery to be tested has a fault.

2. The battery fault identification method according to claim 1, characterized in that: The step of converting the impedance spectrum data from the frequency domain to the time domain to obtain time domain impedance spectrum data comprises: Performing discrete modeling on the impedance spectrum data in combination with a relaxation time distribution model to construct a linear equation; Solving the linear equation to obtain a relaxation time distribution vector; Based on the relaxation time distribution vector, the impedance response characteristics on the time domain scale are inferred, and the time domain impedance spectrum data is obtained based on the impedance response characteristics.

3. The method according to claim 1, characterized in that The step of enhancing the characteristics of the time-domain impedance spectrum data to obtain enhanced impedance spectrum data includes: Construct a preset direction vector; The time-domain impedance spectrum data is projected onto the preset direction vector to obtain the enhanced impedance spectrum data.

4. The method according to claim 3, characterized in that The step of constructing a preset direction vector includes: Obtain impedance spectrum samples; Calculating the distance between each impedance spectrum sample and the classification hyperplane, and adding the reciprocals of the distances to obtain the objective function; Solve the objective function to obtain the preset direction vector.

5. The method according to claim 1, wherein After obtaining the impedance spectrum data, the method further includes: Preprocessing the impedance spectrum data; the preprocessing includes denoising, normalization and standardization; The preprocessing of the impedance spectrum data is obtained by a pre-trained preprocessing model; the training process of the preprocessing model includes: Obtaining first impedance spectrum training samples of multiple labeled state categories; adding preset noise to each of the first impedance spectrum training samples; Calculating the frequency attention weight of the first impedance spectrum training sample after adding random noise at each frequency point, and performing weighted processing on the impedance information of each frequency point according to the frequency attention weight to obtain a weighted impedance spectrum feature; Inputting the weighted impedance spectrum feature into an initial encoder to extract a first feature vector; The first feature vector is input into an initial decoder for impedance reconstruction to obtain a reconstruction result, and the parameters of the initial decoder and the initial encoder are adjusted according to the error between the first impedance spectrum training sample and the reconstruction result until the training is completed to obtain the preprocessing model.

6. The method according to claim 1, characterized in that The training process of the fault identification model includes: Obtaining a second impedance spectrum training sample; wherein the second impedance spectrum training sample carries a fault label; Inputting the second impedance spectrum training sample into the initial model for training to obtain a fault prediction result; A target loss function is obtained based on the predicted fault result and the fault label, and the parameters of the initial model are adjusted according to the target loss function until the training is completed to obtain the fault recognition model.

7. A battery fault identification device, characterized in that: The device comprises: An acquisition module is used to acquire impedance spectrum data; the impedance spectrum data is obtained by applying preset excitations at multiple test frequency points to the battery under test in a working state; A conversion module, configured to convert the impedance spectrum data from the frequency domain to the time domain to obtain time domain impedance spectrum data; A feature enhancement module, configured to enhance the features of the time-domain impedance spectrum data to obtain enhanced impedance spectrum data; The identification module is used to input the enhanced impedance spectrum data into a pre-trained fault identification model to obtain a battery fault identification result; the battery fault identification result is used to indicate whether the battery to be tested has a fault.

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